# Your FREE Guide to Mastering Legal Prompt Engineering with ChatGPT The most in-depth guide on effective prompt engineering specifically for paralegals, attorneys, researchers, and other legal professionals. * **Targeted Responses:** Craft prompts that generate highly relevant AI outputs. * **Quality Control:** Guarantee consistent results and maintain the highest standards in AI-generated outputs. * **Interactive Learning:** Enhance AI's ability to provide insightful, context-aware responses. * **Customized Outputs:** Tailor AI-generated content to suit specific client needs or cases. * **Adaptability:** Train AI to handle diverse or evolving legal scenarios effectively. ## 🚨 We update this guide regularly! 🚨 Get notified when we release new chapters, sections, and example prompts. Fill out the form below to subscribe for notifications. {% embed url="" %} ## Our Sponsors
https://www.a4l.law/Association for AI in Legal/files/AgbyBiV50vWhKycGRUMf
# 1. Introduction - The Power of Precision in Prompting {% hint style="info" %} **Prerequisites:** If you want to follow along with this guide or apply the concepts we cover to create your own legal prompts, make sure you take a look at our prerequisite guide before continuing. {% endhint %} ### What is AI Prompting? Imagine you're delegating a complex legal research task to a junior associate. The quality of their work depends entirely on how clearly you communicate your expectations, the context you provide, and the specific instructions you give. AI prompting works the same way—except your associate is a Large Language Model (LLM) like ChatGPT, Claude, or other generative AI tools. **Prompting** is the art and science of crafting instructions that guide AI systems to produce useful, accurate, and professionally appropriate legal work product. A well-crafted prompt transforms AI from a generic text generator into a powerful legal assistant capable of drafting contracts, analyzing case law, summarizing depositions, and supporting trial strategy. Think of generative AI and LLMs as advanced research and drafting *assistants*, not substitutes for human judgment. These tools are trained on vast amounts of text and can generate human-like responses based on the prompts you provide. However, they don't "understand" law the way you do—they recognize patterns and predict likely continuations of text based on their training data. **Prompt Engineering** is the specialized skill of communicating precisely with AI to receive actionable, verified legal work product. It's the difference between asking "What's summary judgment?" and crafting a detailed instruction that yields a memorandum-ready analysis of summary judgment standards in your specific jurisdiction, applied to your client's fact pattern. ### Why Legal Professionals Need This Guide The legal profession is at a critical inflection point. Generative AI has moved from experimental technology to practical tool faster than any technology in recent history. Yet most legal professionals are still using AI the way most people use search engines in 1998—they know it exists, they've tried it a few times, but they're not unlocking its full potential. #### The Opportunity AI offers legal professionals unprecedented capabilities: * **Time Savings**: Tasks that once took hours can be completed in minutes. A paralegal reviewing 50,000 documents for privilege can use AI to reduce the review set to 5,000 documents requiring human attention—saving hundreds of billable hours. * **Enhanced Accuracy**: AI doesn't get tired, doesn't overlook details due to fatigue, and can maintain consistency across massive document sets. It can catch the synonym you didn't consider when building your search terms. * **Cost Efficiency**: Clients increasingly demand value-based billing. AI allows you to deliver high-quality work faster and at lower cost, improving client satisfaction while maintaining profitability. * **Competitive Advantage**: Firms and practitioners who master AI tools now will have significant advantages in efficiency, quality, and client service. Those who delay risk being left behind in an increasingly AI-enhanced legal marketplace. #### The Challenge However, AI in legal practice comes with serious responsibilities and risks: * **Hallucination**: AI can confidently cite non-existent cases, misstate legal principles, or fabricate facts. Attorneys have been sanctioned for filing briefs with AI-generated fake citations. * **Confidentiality**: Inputting client information into the wrong AI platform can breach attorney-client privilege and violate professional ethics rules. * **Competence**: Model Rule 1.1 requires lawyers to understand the benefits and risks of relevant technology. Using AI without understanding its limitations violates your professional duties. * **Quality Control**: AI output is always a draft requiring human review, verification, and professional judgment. There are no shortcuts around this requirement. This guide bridges the gap between AI's potential and its responsible use in legal practice. You'll learn not just *how* to prompt AI, but *when* to use it, *how* to verify its output, and *how* to integrate it ethically into your practice. ### How to Use This Guide This guide is designed to be both a comprehensive resource and a practical reference you can return to repeatedly. Here's how to get the most value: #### For Different Reader Types **If you're new to AI**: Start with Chapter 2 to understand the fundamentals of prompt engineering. The concepts build progressively, so working through the chapters in order will give you the strongest foundation. **If you have AI experience**: You might skim Chapter 2 and dive directly into Chapter 3's practical examples. Use the prompt templates as starting points and customize them for your specific needs. **If you're a solo practitioner or small firm attorney**: Pay special attention to cost optimization strategies throughout the guide and the workflow integration sections in Chapter 5. **If you're a paralegal or legal assistant**: Focus on the document review, discovery, and administrative task examples. Understanding verification requirements in Chapter 4 is critical for your work. **If you're at a large firm**: The workflow integration strategies in Chapter 5 and the discussion of legal-specific platforms will be particularly relevant for your practice. #### Learning Approach **Hands-On Practice**: The best way to learn prompting is by doing. As you read each chapter, try the examples with your own cases (remembering to redact confidential information). Experiment with variations. See what works and what doesn't. **Iterative Refinement**: Prompt engineering is inherently iterative. Your first prompt rarely produces perfect results. You'll learn to refine prompts based on the AI's responses, gradually steering it toward better output. **Build Your Library**: As you develop effective prompts for common tasks, save them. Create a personal prompt library you can reference and reuse. Many of the examples in this guide are designed as starting templates you can customize. **Stay Current**: AI technology evolves rapidly. While this guide focuses on evergreen principles that will remain relevant, specific AI capabilities continue to advance. Follow updates from major AI providers and legal tech publications. #### Important: What This Guide Is Not This guide teaches you how to use AI effectively and responsibly. It does **not**: * Replace legal research databases or primary sources * Eliminate the need for human judgment and verification * Guarantee that AI output is accurate or error-free * Provide legal advice about AI use in specific situations * Substitute for your law firm's AI usage policies Always verify AI-generated content, follow your firm's technology policies, and consult with risk management when questions arise about AI use in sensitive matters. ### What to Expect: From Basics to Advanced Applications #### Chapter 2: Fundamentals of Legal Prompt Engineering You'll learn the foundational concepts that make the difference between frustrating AI interactions and powerful legal assistance. We cover: * The three golden rules of effective prompting * Common mistakes that lead to poor AI output and how to avoid them * The C.A.S.E. Framework—a structured approach to every legal prompt * How to prevent AI "hallucination" (fabricated facts and citations) * The "Prompt Sandwich" structure for consistent results By the end of Chapter 2, you'll understand why some prompts produce excellent results while others fail, and you'll have a reliable framework for crafting effective prompts every time. #### Chapter 3: Practical Prompting Techniques and Real-World Applications This is where theory meets practice. Chapter 3 provides detailed, copy-and-use prompt examples organized by legal task: * **Discovery and Document Review**: From privilege logs to deposition summaries to exhibit tagging * **Legal Research and Analysis**: Statutory comparison, case law synthesis, jurisdiction-specific research * **Drafting and Client Communication**: Discovery requests, demand letters, client emails, negotiation strategies * **Trial Preparation**: Witness preparation, exhibit cross-referencing, motion drafting Each example includes the prompt structure, explains why it works, and shows you how to adapt it to your specific needs. These aren't generic templates—they're battle-tested prompts designed specifically for legal work. #### Chapter 4: Ethical Guardrails and Professional Responsibility AI isn't just a technical tool—it's a professional responsibility issue. Chapter 4 ensures you use AI in compliance with the Rules of Professional Conduct: * Your non-delegable duty to verify all AI output * How to protect client confidentiality when using AI tools * Understanding and managing AI hallucinations in legal work * The Mandatory AI Review Protocol (MARP) for quality assurance * Candor obligations when using AI-assisted research * Practical compliance frameworks for daily use This chapter also addresses real disciplinary cases where lawyers misused AI, helping you avoid career-damaging mistakes. #### Chapter 5: Building Effective AI Workflows Once you understand the fundamentals and ethics, Chapter 5 shows you how to integrate AI systematically into your practice: * Mapping AI use to each phase of litigation (from early case assessment through trial) * Multi-step workflows that use AI strategically at different stages * Prompt chaining techniques for complex analysis * Quality assurance systems that catch errors before they become problems * Cost management strategies and ROI calculations * Team training and firm-wide implementation This chapter transforms AI from an occasional experiment into a reliable component of your legal practice. #### Chapter 6: Resources and Tools The final chapter provides ongoing reference materials: * Directory of AI platforms (general-purpose and legal-specific) * Curated prompt template library organized by task type * Research papers and further reading * Quick reference guides you can print and keep at your desk * Links to updated resources as AI technology evolves ### A Note on Confidentiality Throughout this guide, all examples use fictional scenarios, generic client names, or publicly available information. When you practice the techniques in this guide: **Never input actual client names, case details, privileged communications, or confidential information into public AI platforms.** Use firm-approved, secure AI tools for work involving client information, or carefully redact and anonymize before practicing with public AI tools. Chapter 4 provides detailed guidance on this critical requirement. ### Let's Begin Mastering AI for legal work isn't about replacing your expertise—it's about amplifying it. The lawyers and paralegals who thrive in the next decade will be those who combine deep legal knowledge with effective AI collaboration. You're about to learn a skill that will transform how you practice law. Let's get started. *** *In the next chapter, we'll dive into the fundamentals of legal prompt engineering, starting with the three golden rules that separate effective prompts from ineffective ones.* # 2. Fundamentals of Legal Prompt Engineering ### The Three Golden Rules of Effective Prompting Before diving into complex frameworks and techniques, you need to understand three foundational principles that separate effective prompts from ineffective ones. These golden rules apply to every prompt you'll ever write, from simple questions to complex multi-step legal analysis. #### Rule 1: Clarity Your prompt should be unambiguous and straightforward. AI models interpret your instructions literally—if there's room for multiple interpretations, you'll get unpredictable results. **Poor Example:** ``` Tell me about summary judgment. ``` **Better Example:** ``` Explain the legal standard for summary judgment in federal court under Rule 56 of the Federal Rules of Civil Procedure, including the burden of proof and the standard for viewing evidence. ``` The second prompt removes ambiguity. Are you asking for a definition? A history? Application to a specific case? The clearer your request, the more useful the response. #### Rule 2: Specificity General prompts produce general responses. Specific prompts produce actionable work product. The more specific you are about what you need, the more the AI can tailor its response to your exact requirements. **Poor Example:** ``` Draft a contract. ``` **Better Example:** ``` Draft a commercial lease agreement for retail space in California, including provisions for: (1) 5-year term with two 3-year renewal options, (2) triple-net lease structure, (3) percentage rent clause tied to gross sales, (4) tenant improvement allowance, and (5) standard force majeure provisions. ``` Specificity transforms AI from a generic tool into a precision instrument. #### Rule 3: Context Context helps the AI understand not just what you're asking, but why you're asking it and how the answer will be used. This shapes the tone, depth, and format of the response. **Poor Example:** ``` What are the elements of negligence? ``` **Better Example:** ``` I am preparing jury instructions for a premises liability case in Texas state court. What are the elements of negligence that must be proven, and how should they be explained to a jury in plain language? ``` The context tells the AI this needs to be jury-appropriate language in a specific jurisdiction, not an academic treatise on tort law. ### Common Pitfalls and How to Avoid Them Understanding what doesn't work is just as important as knowing what does. Here are the most common mistakes legal professionals make when prompting AI. #### Pitfall 1: Vague and Short Statements Most people prompt AI the way they use search engines—short queries that lack detail. **Example of the Problem:** ``` Draft an operating agreement for a private trust company domiciled in Georgia. ``` This prompt seems clear, but it's missing critical information: * Member-managed or manager-managed? * Which Georgia (state, country, city)? * Who are the members/managers? * What is the trust company's purpose? * Is the trustee role bifurcated (administrative vs. distribution)? * Is there an investment committee? **The Solution:** Add layers of specificity and context: ``` Draft a member-managed operating agreement for a private trust company domiciled in the State of Georgia. The trust company has three members: John Doe (50% interest), Jane Smith (30% interest), and ABC Family Trust (20% interest). The company will serve as trustee for family trusts holding both traditional securities and digital assets. Include provisions for: (1) member voting rights proportional to ownership, (2) annual distributions based on company profits, (3) fiduciary duties specific to trust administration, and (4) procedures for adding/removing members. ``` #### Pitfall 2: Lack of Contextual Information Without context, AI makes assumptions that may not align with your needs. **Example of the Problem:** ``` Draft an operating agreement for a private trust company domiciled in Georgia. ``` This prompt fails to mention that the trust company will serve as trustee to specific types of trusts with specific assets. **The Solution:** Provide the relevant contextual backdrop: ``` Draft an operating agreement for a private trust company domiciled in Georgia. The trust company will serve as trustee to an Irrevocable Self-Settled Spendthrift Trust which will hold title to equities and digital assets (cryptocurrency). The trust will follow a HEMS (Healthcare, Education, Maintenance, & Support) distribution model. The operating agreement should address the unique considerations of serving as trustee for trusts holding digital assets. ``` #### Pitfall 3: Not Accounting for Hallucination AI can fabricate information that sounds plausible but is entirely false. In legal work, this is particularly dangerous. **Example of the Problem:** ``` Draft an operating agreement for a private trust company domiciled in Georgia. ``` Here's the fundamental issue: Georgia does not have state legislation recognizing private trust companies. The AI will not alert you to this fact and will proceed to generate an operating agreement anyway. **The Solution:** Build verification requirements into your prompts: ``` You are an estate planning attorney drafting trust agreements and legal contracts. Your work is strictly based on state statutes and federal laws. When asked to draft legal contracts that are state-specific, make sure to reference the state's statutes for a regulatory framework specifically designed for that type of legal contract. If such state statutes do not exist, reply by saying "I cannot complete this request due to regulatory uncertainty" and nothing else. Now, draft an operating agreement for a private trust company in the state of Georgia. ``` With these instructions, the AI correctly responds: "I cannot complete this request due to regulatory uncertainty." #### Pitfall 4: Lack of Structure Unstructured prompts produce inconsistent results. Adding structure to your prompts dramatically improves output quality and reliability. ### The C.A.S.E. Framework The C.A.S.E. Framework is your systematic approach to crafting effective legal prompts. Every successful legal prompt contains these four crucial elements: #### C - Context Define the subject matter and provide background on the task, jurisdiction, court (if applicable), and area of law. Context shapes how the AI interprets and responds to your request. **Elements to Include:** * Subject matter and background * Jurisdiction (federal, state, specific court) * Area of law * Input data references **Example:** ``` I am preparing for a civil trial in the Southern District of New York concerning a breach of commercial lease agreement. The lease was executed in 2020 for retail space in Manhattan. The landlord is claiming $500,000 in damages for early termination. ``` #### A - Audience & Action (Persona) Assign the AI a specific role or persona. This shapes the tone, expertise level, and approach of the response. Then define the specific action you need performed. **Persona Examples:** * "Act as a litigation paralegal preparing document summaries..." * "Assume the role of opposing counsel evaluating weaknesses in my case..." * "You are a senior partner specializing in tort law reviewing an associate's work..." **Action Verbs to Use:** * Summarize * Draft * Compare and contrast * Generate objections * Outline * Analyze * Extract * Identify **Example:** ``` Act as a litigation paralegal preparing for trial. Review the following deposition transcript and generate a table with three columns: (1) Page/Line number, (2) Key Statement, and (3) Inconsistency (compared to the witness's prior statements). ``` #### S - Structure & Style Define exactly how you want the response organized and what tone it should adopt. This ensures the output matches your specific needs and professional standards. **Format Specifications:** * "Provide the summary as a three-column table" * "Format the response as bullet points" * "Draft as a formal email" * "Use only IRAC structure (Issue, Rule, Analysis, Conclusion)" * "Present as a two-page executive summary followed by detailed appendices" **Tone Specifications:** * "Highly persuasive and aggressive" * "Neutral and objective" * "Plain language suitable for a client with no legal background" * "Professional but empathetic" * "Academic and scholarly" **Example:** ``` Format your response as a formal memorandum with the following sections: 1. Summary (3 sentences maximum) 2. Background (chronological narrative) 3. Legal Analysis (using IRAC format) 4. Recommendation (numbered list of next steps) Use professional legal terminology but ensure clarity for a business client. ``` #### E - Ethical and Verification Directives Always include instructions that promote accuracy and identify limitations. This is your safeguard against hallucination and unreliable output. **Citation Requirements:** ``` Where citing case law, provide the full Bluebook citation. For every legal proposition, cite to the primary source (statute, case, regulation). ``` **Limitation Acknowledgment:** ``` If you cannot find a relevant source, explicitly state "I could not locate relevant authority on this issue" rather than fabricating information. If a legal conclusion is speculative or uncertain, clearly indicate this with phrases like "the law is unclear on this point" or "courts have reached conflicting conclusions." ``` **Example of Complete Ethical Directive:** ``` For all legal citations: 1. Provide full Bluebook citations 2. If you're uncertain whether a case exists, state "I recommend verifying this citation in Westlaw or Lexis" 3. Do not cite cases you cannot verify 4. If the law is unsettled, acknowledge conflicting authorities 5. Clearly distinguish between majority and minority positions ``` ### The "Prompt Sandwich" Structure The "Prompt Sandwich" is a proven template that incorporates all elements of the C.A.S.E. Framework in a structured, repeatable format: ``` ┌─────────────────────────────────┐ │ INSTRUCTIONS (Top Bun) │ ← Persona, ethical directives, │ │ general guidelines ├─────────────────────────────────┤ │ CONTEXT (The Filling) │ ← Background, jurisdiction, │ │ relevant facts ├─────────────────────────────────┤ │ INPUT (More Filling) │ ← The specific task, question, │ │ or document to analyze ├─────────────────────────────────┤ │ OUTPUT (Bottom Bun) │ ← Format requirements, structure │ │ specifications, what to exclude └─────────────────────────────────┘ ``` #### Example: Complete Prompt Sandwich ``` **INSTRUCTIONS** You are an estate planning attorney drafting complex trust agreements and other legal contracts. Your work is strictly based on state statutes and federal laws. When asked to draft legal contracts that are state-specific, make sure to reference the state's statutes for a regulatory framework specifically designed for the type of legal contract. If such state statutes do not exist, reply by saying "I cannot complete this request due to regulatory uncertainty" and nothing else. Only include the resulting output in your response. Exclude legal disclaimers, preamble, and any reasoning in your response. **CONTEXT** Client is John Doe of Atlanta, Georgia. [Attach: Irrevocable Trust document] **INPUT** Draft an operating agreement for a private trust company in the state of Nevada that will serve as trustee for the attached irrevocable trust. The trust company will be manager-managed with two managers: John Doe and Jane Doe. The company will have authority to hold and manage both traditional securities and digital assets. **OUTPUT** Format as a professional operating agreement with: - Article I: Organization and Purpose - Article II: Management Structure - Article III: Capital Contributions - Article IV: Distributions - Article V: Fiduciary Duties - Article VI: Amendment Procedures Use formal legal language appropriate for filing with the Nevada Secretary of State. ``` ### Preventing AI Hallucination in Legal Work Hallucination—when AI fabricates information that sounds plausible but is false—poses the greatest risk to legal professionals. Here are strategies to minimize this risk: #### Strategy 1: Explicit Uncertainty Instructions ``` If you do not know the answer or do not have sufficient information, respond by saying "I do not have enough information to answer this question reliably" instead of generating a response. ``` #### Strategy 2: Request Source Attribution ``` For every factual statement, indicate the source of that information. If the information comes from your training data, say "Based on general legal principles" rather than citing to specific authority you cannot verify. ``` #### Strategy 3: Confidence Levels ``` For each conclusion you reach, indicate your confidence level: - HIGH CONFIDENCE: Well-established legal principle with clear authority - MEDIUM CONFIDENCE: Generally accepted principle but with some variation - LOW CONFIDENCE: Uncertain or evolving area of law - SPECULATIVE: No clear authority; this is an educated inference ``` #### Strategy 4: Multiple Verification Steps Don't rely on a single AI response. Use this multi-step verification approach: 1. **Initial prompt** with strong verification requirements 2. **Second prompt** asking the AI to identify weaknesses or uncertainties in its first response 3. **Manual verification** of all citations and key legal propositions in primary sources **Example Second-Step Prompt:** ``` Review your previous response. Identify any statements that you are not highly confident about. For each citation provided, indicate whether you are certain this case exists and accurately supports the proposition cited. ``` ### Putting It All Together: Before and After Examples #### Example 1: Contract Review **Poor Prompt:** ``` Review this contract and tell me if there are any issues. ``` **Improved Prompt Using C.A.S.E.:** ``` **INSTRUCTIONS** Act as a senior corporate attorney conducting due diligence review. Focus on risk identification and business impact. For any issue identified, assess the severity as Critical, High, Medium, or Low. **CONTEXT** I represent the buyer in an asset purchase transaction valued at $5 million. We are acquiring a software-as-a-service business. This is a key supplier agreement that will transfer to us post-closing. **INPUT** Review the attached Master Services Agreement between the target company and their primary cloud infrastructure provider. Identify and analyze: 1. Termination provisions and change of control implications 2. Liability limitations and indemnification 3. Data security and privacy obligations 4. Pricing and renewal terms 5. Any provisions that could create post-closing issues **OUTPUT** Provide response in this format: EXECUTIVE SUMMARY (3-4 sentences) CRITICAL ISSUES (if any) - Issue description - Business impact - Recommended action HIGH-PRIORITY ISSUES [Same format as above] MEDIUM/LOW-PRIORITY ISSUES [Same format as above] OVERALL RISK ASSESSMENT [One paragraph summary] ``` #### Example 2: Legal Research **Poor Prompt:** ``` What's the law on non-compete agreements in California? ``` **Improved Prompt Using C.A.S.E.:** ``` **INSTRUCTIONS** You are a legal research specialist. Provide comprehensive analysis with full citations. If any statement is based on interpretation rather than explicit statutory or case law, clearly indicate this. If the law has changed recently or is subject to pending legislation, note this. **CONTEXT** I represent an employer who is considering requiring new employees to sign non-compete agreements. The company is headquartered in California but has employees in several states. We need to understand California's approach before rolling out any policies. **INPUT** Research and analyze California law regarding the enforceability of employee non-compete agreements. Address: 1. The general rule in California regarding non-competes 2. Statutory basis (cite specific California statutes) 3. Key exceptions or narrow circumstances where non-competes may be enforceable 4. Recent case law developments (last 5 years) 5. Practical alternatives employers can use to protect business interests **OUTPUT** Structure your response as: OVERVIEW (one paragraph summary of California's position) STATUTORY FRAMEWORK - Cite relevant California Business & Professions Code sections - Explain key statutory language EXCEPTIONS AND SPECIAL CASES - Business sale exception - Trade secret protection - Other recognized exceptions CASE LAW ANALYSIS - 2-3 key cases with full citations - Brief summary of holdings PRACTICAL ALTERNATIVES - Non-solicitation agreements - Confidentiality agreements - Other protective measures All case citations must include: Case name, citation, court, and year. Example: Edwards v. Arthur Andersen LLP, 44 Cal. 4th 937 (2008) ``` ### Practice Exercise To reinforce these concepts, take this poorly constructed prompt and rebuild it using the C.A.S.E. Framework: **Poor Prompt:** ``` Summarize this deposition. ``` **Your Task:** Before looking at the answer below, try rewriting this prompt to include: * Clear context about the case and the deposition's significance * A specific persona for the AI * Detailed structure requirements * Ethical verification requirements **Example Improved Version:** ``` **INSTRUCTIONS** Act as a litigation paralegal preparing materials for trial counsel. Focus on identifying admissions, inconsistencies with prior testimony, and statements that support or undermine our case theory. Flag any testimony that may require follow-up in future depositions. **CONTEXT** This is a products liability case involving an allegedly defective medical device. The plaintiff claims the device malfunctioned during surgery, causing permanent injury. This deposition is of the defendant's engineering director who oversaw the device's design and testing. Our case theory: The device had a known design flaw that the defendant failed to disclose. Defendant's position: The device functioned properly and any injury was due to surgical error. **INPUT** Summarize the attached 200-page deposition transcript of Dr. Robert Chen, Engineering Director. Focus specifically on: 1. Testimony regarding the device's testing protocols 2. Any admissions about design modifications or safety concerns 3. What Dr. Chen knew about prior incidents with the device 4. Statements that contradict the company's public statements or marketing materials **OUTPUT** Provide a summary in this format: KEY ADMISSIONS (most important testimony supporting our case) - Quote with page:line citation - Significance for our case theory INCONSISTENCIES (testimony that conflicts with other evidence) - Statement from this deposition (page:line) - Conflicting prior statement (source and citation) - Potential impact HELPFUL DEFENSE TESTIMONY (testimony that supports defendant's case) - Quote with citation - How this may be used against us AREAS FOR FOLLOW-UP (topics needing clarification or further inquiry) CREDIBILITY ASSESSMENT (witness's demeanor, evasiveness, areas of uncertainty) Maximum 5 pages. Use bullet points. Include specific page and line citations for every quoted statement. ``` ### Chapter Summary Mastering legal prompt engineering begins with understanding and applying these fundamental principles: 1. **Follow the Three Golden Rules**: Every prompt must be clear, specific, and contextual 2. **Avoid Common Pitfalls**: Vague prompts, missing context, and unstructured requests produce poor results 3. **Use the C.A.S.E. Framework**: Context, Audience & Action, Structure & Style, and Ethical directives 4. **Apply the Prompt Sandwich Structure**: Organize your prompts systematically for consistent results 5. **Prevent Hallucination**: Build verification requirements into every prompt These fundamentals will serve as the foundation for everything that follows. In the next chapter, we'll apply these principles to real-world legal tasks, providing you with ready-to-use prompt templates for discovery, research, drafting, and trial preparation. *** *In Chapter 3, we'll move from theory to practice with detailed examples of prompts for every stage of litigation—from early case assessment through trial preparation.* # 2.1. Prompt Templates We briefly introduced prompt templates in [chapter 1](/1.-introduction-the-power-of-precision-in-prompting) when describing how you can use the "Prompt Sandwich" template to get better results from ChatGPT. In this section we dive deeper into other prompt templates you can use in different scenarios to get better results. ## Templates by Common Task Types: {% content-ref url="/pages/aVVsKJz47ploiaZQdbGv" %} [2.1.1. Summarization](/2.-fundamentals-of-legal-prompt-engineering/2.1.-prompt-templates/2.1.1.-summarization) {% endcontent-ref %} {% content-ref url="/pages/oEL5vt0W9a1Unv890PwB" %} [2.1.2. Classification](/2.-fundamentals-of-legal-prompt-engineering/2.1.-prompt-templates/2.1.2.-classification) {% endcontent-ref %} {% content-ref url="/pages/7WoIWr5IPqBwXDx0KePH" %} [2.1.3. Extraction](/2.-fundamentals-of-legal-prompt-engineering/2.1.-prompt-templates/2.1.3.-extraction) {% endcontent-ref %} # 2.1.1. Summarization LLMs like ChatGPT are really good at summarizing information for us. To get the most out of having LLMs summarize large volumes of text, you can provide a template that not only details your desired formatting, but also includes placeholders for the types of information you want included. Let's take a look at a simple example regarding the Securities Act of 1933 starting with your basic prompt asking for a summary: ``` Basic Prompt W/O Template: Summarize the Securities Act of 1933 ``` When given this prompt, ChatGPT will do an okay job at producing at a summary that is legible and informational. Take a look at the response we received [here](https://chat.openai.com/share/9dac8f14-e1f1-4082-bd6d-13d4bec30322). How can we get ChatGPT to generate a better summary? We can use a prompt template that better instructs the AI on the specific types of information you're interested in (dates, persons, etc.), formatting preferences, desired length, and more. Take a look at the following prompt: 
Detailed Summary Prompt Template:
As a legal professional, analyze the securities act of 1933 and generate a 
detailed summary with 3-4 highlights for each of the most important sections 
with important keywords, people, numbers, and facts in this format:

# {title here}

### {section title here}

{START_DETAILS_SECTION}
{summary of the section with important keywords, people, numbers, and facts}

- {first point}: {short explanation with important keywords, people, and facts}
- {second point}: {same as above}
- {third point}: {same as above}
<!-- a fourth point if warranted, and so on -->
{END_DETAILS_SECTION}

### {second section here}

{START_DETAILS_SECTION}
{summary of the section with important keywords, people, numbers, and facts}

- {first point}: {short explanation with important keywords, people, and facts}
- {second point}: {same as above}
- {third point}: {same as above}
<!-- a fourth point if warranted, and so on -->
{END_DETAILS_SECTION}

### {third section here}
<!-- and so on, as many sections and details/summary subpoints as warranted -->

End with other questions that the user might want answered based on this source:

### Further information
- {question + answer 1}
- {question + answer 2}
- {question + answer 3}

All words in brackets must replaced by the summary of the content.
Only draw from the source content, do not hallucinate.

Only output the response in the prescribed format with NO additional commentary. 
Answer the questions included in the "Further information" section and display 
the answer alongside the question.
Follow along with us as we prompt ChatGPT using this template [here](https://chat.openai.com/share/3346e40f-5f78-4a1e-adce-c1ab3174f6e7). There is a lot to digest in this prompt template, but for starters don't shy away from the weird syntax being used. This syntax is called [Markdown](https://en.wikipedia.org/wiki/Markdown) and can be used to format raw text using headings, lists, and other elements you are used to seeing in Microsoft Word and Google Docs. It's great at helping guide the AI model towards producing a response in a prescribed format, but also allows you to embed instructions within as well, a good example being the list of important keywords, people, and facts we want generated for each section as illustrated: ``` ... - {first point}: {short explanation with important keywords, people, and facts} ... ``` Hover over the annotated sections of our improved prompt template above to learn more about the specific characteristics we included in this prompt and how they work together. We encourage you to tinker with this template to see if you can get even more precise summaries based on your particular use case. For example, if your law firm prefers to have case files, or legal analysis drafted in a desired format with specific criteria regarding the information that should be included, go ahead and modify the prompt template and see how ChatGPT responds. Remember, prompt engineering is an iterative process. [^1]: The "#" character is used for document headings. [^2]: This is a placeholder. The AI will replace all text within curly braces ("{}") with generated content. [^3]: "###" is used for page subheadings. [^4]: This special placeholder text informs the AI the beginning of detailed information for a particular section of content. [^5]: Any text found within "\" is a template comment/annotation used to help guide the AI. These are omitted from the model's response. [^6]: This special placeholder text informs the AI the end of the details section. [^7]: We want to prevent the LLM from hallucinating and including information in the summary that is inaccurate or completely fabricated. # 2.1.2. Classification LLMs are good at performing classification tasks, assigning a class or category to text. You provide the AI model with a list of categories to choose from (or the model can generate its own categories) and the content you want analyzed. Text classification serves various applications, including fraud detection, sentiment analysis, and content monitoring, among others, thereby proving to be advantageous for legal professionals. ### Example: Personal Injury Cases We work with several firms specializing in personal injury which spend countless hours producing medical chronology reports and medical summaries. Let's imagine we want craft a prompt that is able to identify a client's diagnosis from the discharge notes found within a client's health records from their visit to the ER following the accident. This is a pretty simple classification problem. Let's see how we might be able to achieve this with the following prompt template: 
Prompt:
Correctly identify the patient's diagnosis from the medical summary provided.

Examples:
Summary: Mr. Doe presented in the ER following a rear-end collision while he was 
stationary at a red light. He reports immediate onset of neck pain and stiffness, 
worsening  over the subsequent hours. Physical examination showed a decreased range 
of motion in the neck.
Diagnosis: Whiplash

Summary: Mr. White presented to the clinic several months following a violent mugging. 
He reports recurring nightmares, flashbacks, and increased anxiety, especially 
in crowded places. Psychiatric evaluation confirmed the diagnosis. 
Diagnosis: Post-traumatic stress disorder.

Summary Ms. Smith was brought into the ER after falling off a ladder at home. 
She reports headache, confusion, and some memory loss regarding the incident. 
Neurological examination revealed mild disorientation.
Diagnosis:
You can see that we've used what we learned earlier in this guide to create a prompt that has three different sections. We start our prompt by including an **instruction** – identify the patient's diagnosis. We then embed two **examples** demonstrating the type of medical summary the AI model can expect. The last section is our **input**, which consists of the medical summary and the output delimiter – "Diagnosis." Inputting this prompt into ChatGPT correctly returns the result: "Concussion" as been seen [here](https://chat.openai.com/share/abd45028-3fd8-4b38-998a-85736ee37b8a). [^1]: Instruction [^2]: Label that marks the beginning of example section. # 2.1.3. Extraction LLMs can also effectively extract information from large volumes of text. Common use cases involving extraction prompts include, but are not limited to: * **Information Retrieval**: Extracting specific information from large datasets or complex documents, like identifying key terms, phrases, or sections in legal documents. * **Named Entity Recognition (NER)**: Extracting named entities such as people, organizations, locations, dates, etc. from a text. * **Event Extraction**: Identifying and extracting key events, participants, etc. from a given text. ### Example: Corporate Organizational Charts The blockchain industry has had a string of tumultuous events shakeup the recently. Arguably, the most infamous event was the collapse of FTX, the once high-flying cryptocurrency exchange and hedge fund that filed for bankruptcy in November of 2022. To be more specific, FTX, and **over 100** of its subsidiaries filed for bankruptcy. Imagine the amount of effort required from bankruptcy lawyers to uncover the intricacies of the relationships between all these corporate entities into something that can be easily understood... Enter ChatGPT 🙂 The example prompt below demonstrates how we can perform NER extraction to compile an organizational chart. ``` Prompt: Identify all corporate entities (businesses), along with their parent companies, subsidiaries, and affiliates in the text below. Output in dot language. Text: Plaintiff Securities and Exchange Commission (the “SEC” or the “Commission”) for its Complaint against Defendants Coinbase, Inc. (“Coinbase”) and Coinbase Global, Inc. (“CGI”) (collectively, “Defendants”). CGI—Coinbase’s parent company to which Coinbase’s revenues flow—is a control person of Coinbase and thus violated the same Exchange Act provisions as Coinbase. ``` This is a contrived example, but it shows that we can direct ChatGPT to extract specific types of information along with additional metadata. Here, we're not simply asking for an exhaustive list of all corporate entities mentioned in the text, we also want to understand the relationships between these entities. You can see our interaction with ChatGPT using this prompt [here](https://chat.openai.com/share/99790c89-a1a5-49ac-bead-48ac3d8689e7). Something that we haven't seen yet, but will be discussed at length in [chapter 4](/4.-ethical-guardrails-and-professional-responsibility) are "Output Parsers." In our example prompt above, we provide ChatGPT with a specific instruction – "Output in dot language." ChatGPT keeps things simple, it receives text as input, and returns text as output. One thing that many users don't realize, is that we can ask ChatGPT to return different types of text, text that has specific syntactical meaning, text that represents a computer programming language, etc. In this example, we're interested in generating some type of visual representation of an organizational chart. In this prompt, we are asking ChatGPT to provide its response in Dot language (used to generate visual diagrams–will be covered in [chapter 4](/4.-ethical-guardrails-and-professional-responsibility)). We can take the Dot language response provided by ChatGPT to generate an image like the one below:

This image was generated via ChatGPT's response:
digraph { "Coinbase Global, Inc." -> "Coinbase, Inc." [label="Parent"] }

Extraction prompts can be used heavily by law firms to solve a variety a different problems. In the next section, we'll present an example that uses extraction with few-example prompting to automate a common workflow dreaded by countless attorneys and paralegals. # 2.2. Few-Example Prompting Extensive research has been done by AI experts on how to get better results from LLMs. One of the things they've spent a lot of time looking into is the idea of providing a few examples within prompts. Can providing examples help guide the understanding of the LLM of the task it's been asked to perform...? Researchers found that in many situations, providing examples within your prompts leads to more accurate results returned from the LLM. In research circles, this approach is more formally referred to as "Few-shot" prompting. We've opted towards referring to this strategy as "few-example" prompting to make it more human readable. At this point, you've seen few-example prompting on numerous occasions in previous sections of this guide. This strategy is somewhat universal when it comes to prompt engineering. Lets dive into some "examples" :joy:. ## Example: Preparing for Depositions You can have ChatGPT generate an *exhaustive* list of questions to ask a witness during a deposition. However, you don't want to ask questions that are not relevant to the case. 
Prompt:
Prepare questions for depositions.

Examples:
Input: Generate questions to ask a witness during a deposition in a car accident case?
Output:
- Can you describe the events leading up to the accident?
- What were the weather and road conditions?
- Did you admit fault or make any statements about the accident at the scene?

Input: Create a list of questions to ask a defendant during a deposition in a 
workplace discrimination case?
Output:
- Are you aware of the company's policies regarding workplace discrimination?
- Did the plaintiff make you aware of the alleged discriminatory behavior? 
- Were any actions taken by the company after the alleged incidents were reported?

Input: Questions to ask a witness during a deposition in a commercial lease dispute?
Output:
If you're curious in following along and seeing ChatGPT's response, take a look at our conversation [here](https://chat.openai.com/share/92fda713-e095-4924-b4e8-dd6259b1f136). ## Example: Case Briefings Supercharge your legal research and case analysis & strategy by providing ChatGPT a few examples. 
Prompt:
Create a case briefing with details on the issue, rule, analysis, and conclusion.

Examples:
Case: "Brown v. Board of Education, 347 U.S. 483 (1954)"
Output:
- Issue: Does segregation of public schools based on race deprive minority children 
of equal protection under the law as guaranteed by the 14th Amendment?
- Rule: The Equal Protection Clause of the 14th Amendment.
- Analysis: The Court found that segregation in public education has a detrimental 
effect on minority children because it is interpreted as a sign of inferiority. 
The impact is greater when it has the sanction of the law.
- Conclusion: The Court held that "in the field of public education the doctrine of 
'separate but equal' has no place," as segregated schools are inherently unequal.
 
Case: "Roe v. Wade, 410 U.S. 113 (1973)"
Output:
Again, we can see [here](https://chat.openai.com/share/2f3d0efa-d9d1-47d4-bf6b-ec93d524bfd9) that ChatGPT provided a response matching the format of the example we provided. ## Example: Medical Chronologies Let's change things up a bit by demonstrating how we can help guide the AI model to return its response in a format we desire by giving it a few examples. This example touches on [output parsers](/4.-ethical-guardrails-and-professional-responsibility/4.2.-output-parsers) covered in chapter 4. Many firms choose to produce medical chronology reports using a spreadsheet. You can import data into Microsoft Excel (or similar software) in a variety of different data formats. One of the more common formats is [comma-separated values](https://en.wikipedia.org/wiki/Comma-separated_values) (CSV). Lucky for us, ChatGPT can work CSV data. Here's an example on how we can have ChatGPT return CSV data that we can then import into our spreadsheet. ``` Prompt: Extract the date, event, and generate a one-sentence description from the provided medical summaries and output in CSV format. Examples: Summary: On July 1, 2023, John Doe was involved in a car accident, during which he reported immediate pain in his neck, back, and left shoulder. Output: "2023-07-01","Accident Occurred","John Doe was involved in a car accident. Reported immediate pain in neck, back and left shoulder." Summary: A day after the accident, on July 2, John had a detailed consultation with an orthopedic specialist. The orthopedist, recognizing the urgency of the situation, reset his shoulder back into place. Despite this progress, John was not out of the woods yet – the orthopedist strongly recommended that he commence physical therapy immediately. Output: "2023-07-02","Orthopedic Visit","Consultation with an orthopedist. Shoulder set back into place. Recommended physical therapy." Summary: John started his physical therapy on July 4. It was a grueling session, with John reporting significant pain throughout the process. Despite the discomfort, John remained committed to the rehabilitation process, understanding that this was a crucial step towards his recovery. Output: ``` Don't worry about trying to interpret all of the quotes and commas in the output. Instead, have ChatGPT display a table for you like the one shown below. Reference [our interaction](https://chat.openai.com/share/2bb8fd0e-6300-4cdc-8bce-2c355a93cd4f) with ChatGPT, and experiment further on how you can generate structured data with small changes to your prompt. ## Summary Few-example prompting is analogous to how humans learn from only a few examples. For instance, if you show a child a few pictures of dogs and tell them "These are dogs," they can typically recognize other dogs they haven't seen before. It's important to note that while few-shot learning can be quite effective, it's not perfect. The quality of responses can vary based on the complexity of the prompt, the quality and relevance of the examples given, and the inherent limitations of the AI model itself. However, this is a valuable low-cost prompting strategy you should be using regularly. [^1]: Instruction [^2]: A few examples you are providing to ChatGPT to help it in providing answers to your future prompt(s). [^3]: This is our response delimiter. We want ChatGPT to generate a response and append it. # 3. Practical Prompting Techniques and Real-World Applications This chapter provides battle-tested prompt templates organized by legal task type. Each section explains the technique, then provides ready-to-use examples you can adapt to your specific needs. Remember: these prompts follow the C.A.S.E. Framework and Prompt Sandwich structure you learned in Chapter 2. ### Discovery and Document Review Discovery is one of the most time-intensive phases of litigation. AI can dramatically accelerate document review, privilege logging, deposition analysis, and exhibit management—but only with properly structured prompts. #### Deposition Summary and Analysis **When to Use**: After receiving a deposition transcript that requires summarization for trial preparation or motion practice. **Key Technique**: Provide clear case theory context so the AI can identify relevant testimony that supports or undermines your position. **Example Prompt:** ``` **INSTRUCTIONS** Act as a litigation paralegal preparing materials for trial counsel. Focus on identifying admissions, inconsistencies with prior testimony, and statements that support or undermine our case theory. Flag any testimony that may require follow-up in future depositions. If you identify inconsistencies, cite both the current testimony and the conflicting prior statement with specific page and line numbers. **CONTEXT** This is an employment discrimination case. Plaintiff alleges she was terminated due to age discrimination. This deposition is of the HR Director who participated in the termination decision. Our case theory: Plaintiff was terminated because of age-based comments made by her supervisor and a pattern of replacing older employees with younger ones. Defendant's position: Plaintiff was terminated due to legitimate performance issues documented in her file. **INPUT** Summarize the attached deposition transcript of Sarah Martinez, HR Director. Focus specifically on: 1. Her knowledge of the supervisor's comments about the plaintiff's age 2. Her involvement in the termination decision-making process 3. Any documentation she reviewed before the termination 4. Her awareness of the company's hiring patterns and employee ages 5. The performance issues she claims justified termination **OUTPUT** Provide a summary in this format: KEY ADMISSIONS (testimony supporting our case) - Direct quote with [page:line] citation - Why this matters for our case theory INCONSISTENCIES (conflicts with other evidence) - Current testimony [page:line] - Prior conflicting statement [source and citation] - Significance DEFENSE-FAVORABLE TESTIMONY - Quote with citation - How defense may use this CREDIBILITY ISSUES - Evasive responses - Areas of uncertainty - Changes in demeanor FOLLOW-UP TOPICS - Issues requiring clarification - Documents to request - Additional witnesses to depose Maximum 5 pages. Include specific page:line citations for all quotes. ``` #### Privilege Log Creation **When to Use**: When conducting document review for privilege determination and need to draft privilege log entries. **Key Technique**: Define specific criteria for privilege while instructing the AI to flag documents requiring human review rather than making final determinations. **Example Prompt:** ``` **INSTRUCTIONS** Act as a discovery coordinator identifying potentially privileged documents for attorney review. Your role is to FLAG documents that may be privileged, not to make final privilege determinations. Be conservative—when in doubt, flag for review. Only exclude documents that clearly have no privilege claim (e.g., purely business emails with no attorney involvement). **CONTEXT** We are conducting document review in a contract dispute. The producing party is XYZ Corporation. Relevant time period: January 1, 2023 to present. Known attorney email domains: - @smithlaw.com (outside counsel) - legal.dept@xyzcorp.com (in-house counsel) **INPUT** Review the attached 500 emails from the "Legal-Comms" folder. For each potentially privileged document, extract: 1. Document ID/Bates number 2. Date 3. Author 4. Recipient(s) 5. Subject line 6. Brief description (1-2 sentences, no privileged content) 7. Privilege type claimed (attorney-client, work product, both) 8. Reason for privilege claim **OUTPUT** Create two lists: PRIVILEGED/FOR REVIEW (documents requiring attorney review) Format as table with columns: | Doc ID | Date | From | To | Subject | Description | Privilege Type | Basis | CLEARLY NOT PRIVILEGED (can be produced) - Count only, do not list individually FLAG FOR SPECIAL REVIEW (uncertain or complex privilege issues) - List with explanation of the privilege question presented Total: [X] documents reviewed, [Y] flagged as potentially privileged, [Z] clearly not privileged ``` #### Exhibit Tagging and Organization **When to Use**: When preparing exhibits for trial or organizing discovery productions. **Key Technique**: Create consistent categorization schemes that align with your case organization system. **Example Prompt:** ``` **INSTRUCTIONS** Act as a trial preparation specialist organizing exhibits. Extract key metadata to enable efficient exhibit management and cross-referencing. Be precise with dates—use MM/DD/YYYY format. If a date is unclear or ambiguous, note "DATE UNCERTAIN" and explain the issue. **CONTEXT** Products liability case involving defective automotive parts. We represent the plaintiff. Trial is scheduled for six months from now. Exhibit categories: - DESIGN: Design documents, specifications, engineering drawings - TESTING: Test results, safety analyses, quality control records - COMMS: Internal communications about the product - INCIDENTS: Prior complaints, incident reports, warranty claims - FINANCIAL: Sales data, cost analyses, financial impact documents - REGULATORY: Government filings, compliance documents **INPUT** Review the attached 50 documents designated as trial exhibits. For each document, extract: 1. Exhibit number (if assigned, otherwise note "UNASSIGNED") 2. Document type (email, report, memo, spreadsheet, etc.) 3. Date of document 4. Author/sender 5. Recipient(s) (if applicable) 6. Primary category (from list above) 7. Secondary category (if applicable) 8. Key search terms (3-5 relevant keywords or names) 9. One-sentence summary 10. Relationship to key issues (if apparent) **OUTPUT** Format as a structured table: | Ex. No. | Doc Type | Date | Author | To | Category | Keywords | Summary | Key Issues | After the table, provide: CROSS-REFERENCE NOTES - Documents that reference other exhibits - Chronological clusters (groups of related documents) - Gaps in the documentary record TRIAL PRESENTATION RECOMMENDATIONS - Documents requiring enlargement for jury viewing - Documents needing redaction - Complex documents requiring expert explanation ``` #### Technology Assisted Review (TAR) Training **When to Use**: When setting up or refining predictive coding/TAR protocols for large-scale document review. **Key Technique**: Use AI to analyze documents and suggest search terms, but always validate findings against actual review results. **Example Prompt:** ``` **INSTRUCTIONS** Act as an eDiscovery consultant analyzing document review patterns. Your goal is to identify search terms and criteria that will improve the efficiency of our document review. Base suggestions only on the sample documents provided. Do not speculate about documents you haven't seen. **CONTEXT** Securities fraud litigation. We represent defendant investment firm. Discovery focuses on communications about investment recommendations made to plaintiff between 2020-2023. Current search terms are returning too many irrelevant documents. We need to refine our approach. **INPUT** Analyze the attached sample set of 100 documents: - 40 marked RELEVANT by reviewers - 60 marked NOT RELEVANT by reviewers Identify patterns including: 1. Keywords that appear frequently in relevant documents 2. Keywords that appear frequently in irrelevant documents 3. Sender/recipient patterns in relevant vs. irrelevant documents 4. Date ranges with higher concentrations of relevant documents 5. Document types (email, memo, report) most likely to be relevant **OUTPUT** Provide analysis in this format: RECOMMENDED INCLUSION CRITERIA Terms/patterns that increase likelihood of relevance: - Keyword/phrase: [frequency in relevant docs vs. irrelevant docs] - Explanation of why this matters RECOMMENDED EXCLUSION CRITERIA Terms/patterns that decrease likelihood of relevance: - Keyword/phrase: [frequency data] - Explanation SENDER/RECIPIENT ANALYSIS High-value custodians (appear frequently in relevant documents): - Name, role, frequency Low-value custodians (rarely appear in relevant documents): - Name, role, frequency DATE CLUSTERING Time periods with highest relevance rates: - Date range: [start - end] - Relevance rate: [X%] - Key events during this period REFINED SEARCH STRATEGY Suggested Boolean search string incorporating findings above QUALITY CONTROL RECOMMENDATIONS - Sample size for next training round - Types of documents to prioritize for review ``` ### Legal Research and Analysis AI can accelerate legal research, but verification is critical. These prompts are designed to produce research memoranda that still require attorney validation of all citations and legal conclusions. #### Statutory Comparison Across Jurisdictions **When to Use**: When analyzing how different states or jurisdictions treat the same legal issue. **Key Technique**: Request structured comparison that highlights key differences, not just a narrative description. **Example Prompt:** ``` **INSTRUCTIONS** You are a legal researcher analyzing multi-jurisdictional law. Provide comprehensive analysis with citations to primary sources. For each jurisdiction, cite the specific statute or code section. If you are uncertain about current law, state: "This requires verification in [Westlaw/Lexis]." Do not fabricate citations. If you cannot find a relevant statute, state: "No specific statute located; common law may govern." **CONTEXT** Client is expanding operations to multiple states and needs to understand non-compete agreement enforceability in each jurisdiction. This will inform our employment agreement drafting strategy. **INPUT** Compare non-compete agreement enforceability in the following states: California, Texas, Florida, New York, and Illinois. For each state, analyze: 1. General rule (enforceable, unenforceable, or conditional) 2. Statutory framework (specific code sections) 3. Required elements for enforceability (if applicable) 4. Time limitations (maximum duration enforced) 5. Geographic scope limitations 6. Protectable interests recognized 7. Special rules for specific industries or employee types 8. Recent legislative changes (last 5 years) 9. "Blue pencil" or reformation availability **OUTPUT** Structure your response as follows: EXECUTIVE SUMMARY One-paragraph overview of the spectrum from most to least restrictive. JURISDICTION-BY-JURISDICTION ANALYSIS [For each state:] STATE NAME General Rule: [one sentence] Statutory Cite: [specific code section with full citation] Key Elements: [numbered list] Time Limits: [specific maximum, if any] Geographic Limits: [specific restrictions, if any] Protectable Interests: [what employers can protect] Special Rules: [industry-specific or employee-specific rules] Recent Changes: [legislative updates since 2019] Reformation: [yes/no and explanation] Practical Effect: [how this plays out in practice] COMPARATIVE MATRIX Create a table comparing all five states across key factors: | Factor | CA | TX | FL | NY | IL | STRATEGIC RECOMMENDATIONS - Most employer-friendly jurisdiction for our client - Least employer-friendly jurisdiction - Jurisdictional drafting considerations - Choice of law clause recommendations VERIFICATION NOTE List any citations or legal propositions that require verification in Westlaw or Lexis before relying on this analysis. ``` #### Case Law Synthesis for Motion Practice **When to Use**: When researching case law to support a specific legal argument in motion practice. **Key Technique**: Structure the AI's analysis around your specific argument, not just general legal principles. **Example Prompt:** ``` **INSTRUCTIONS** You are a senior associate preparing a research memorandum for a dispositive motion. Your analysis should be thorough, objective, and cite-checked. For every case cited: 1. Provide full Bluebook citation 2. Include parenthetical explaining the holding 3. Note the procedural posture 4. Flag if the case is binding or persuasive authority If you are uncertain about a case citation or holding, state: "VERIFY: [explanation of uncertainty]" **CONTEXT** We represent defendant in a negligence action arising from a slip-and-fall at a retail store. Plaintiff tripped over merchandise that had fallen into the aisle approximately 30 seconds before the fall. Store employee witnessed the merchandise fall but was assisting another customer and had not yet reached the fallen item. We are filing a Motion for Summary Judgment arguing defendant had no constructive notice of the hazard and insufficient time to remedy it. Jurisdiction: Florida state court (applicable law: Florida premises liability) **INPUT** Research and analyze Florida case law on premises liability, specifically: 1. The "constructive notice" standard in slip-and-fall cases 2. What constitutes "sufficient time" for a property owner to discover and remedy a hazard 3. How courts treat cases where the hazard existed for very brief periods 4. Whether active assistance of other customers affects the duty to remedy hazards 5. The plaintiff's burden in establishing constructive notice Identify: - The leading cases establishing the legal standard - Recent cases applying this standard (last 10 years) - Cases with factually similar circumstances - Any cases that cut against our argument (opposing counsel will cite these) **OUTPUT** Structure as a formal research memorandum: ISSUE PRESENTED [One sentence framing the legal question] BRIEF ANSWER [2-3 sentences: Can we prevail on summary judgment and why?] APPLICABLE LEGAL STANDARD The legal test for constructive notice in Florida premises liability cases: - [Element 1 with citation] - [Element 2 with citation] - [etc.] ANALYSIS A. Constructive Notice Requires Sufficient Time to Discover and Remedy [Discuss leading cases with full citations and analysis] B. Florida Courts Require Evidence of Adequate Time Period [Discuss cases addressing temporal element] C. Brief Duration of Hazard Supports Defendant [Discuss cases with similar short time periods] [Include case comparisons: "Unlike X case where..., here..."] D. Employee's Competing Duties [Discuss whether assisting customers affects notice analysis] CONTRARY AUTHORITY Cases plaintiff will likely cite: - [Case name and citation] - How it differs from our facts - Potential counter-arguments CONCLUSION Summary of likelihood of success on summary judgment based on this research. VERIFICATION CHECKLIST [ ] All case citations checked in Westlaw/Lexis [ ] Shepardizing completed for all cited cases [ ] Procedural posture verified for all cases [ ] Binding vs. persuasive authority confirmed [ ] Recent legislative changes reviewed Cases requiring verification: [list any uncertain citations] ``` #### Regulatory Compliance Analysis **When to Use**: When analyzing whether a client's conduct complies with applicable regulations. **Key Technique**: Request step-by-step analysis that applies regulations to specific facts, not generic compliance advice. **Example Prompt:** ``` **INSTRUCTIONS** You are a regulatory compliance specialist analyzing whether specific conduct complies with federal regulations. Your analysis should be methodical, cite-specific, and identify both clear violations and grey areas. Cite to specific regulatory sections (e.g., "17 CFR § 240.10b-5"). If a regulation is ambiguous or its application to these facts is unclear, state: "AMBIGUOUS: [explanation of the interpretive question]" **CONTEXT** Client is a financial services firm that offers investment advice. We are analyzing whether certain marketing materials and client communications comply with SEC regulations, specifically: - Securities Act of 1933 - Securities Exchange Act of 1934 - Investment Advisers Act of 1940 - Relevant SEC rules and regulations This is a compliance review before launch of new marketing campaign. **INPUT** Analyze the attached marketing materials for regulatory compliance. Review for: 1. Disclosure requirements (mandated disclaimers, risk warnings) 2. Prohibited statements (guaranteed returns, misleading performance data) 3. Required legend or boilerplate language 4. Performance advertising rules (if applicable) 5. Testimonial and endorsement rules 6. Social media-specific requirements (if materials include social posts) 7. Record-keeping requirements triggered by these materials For each potential issue, identify: - The specific regulation implicated - Whether this is a clear violation, grey area, or compliant - The potential penalty or enforcement risk - Recommended corrective action **OUTPUT** Structure as a compliance memorandum: EXECUTIVE SUMMARY Overall compliance assessment: [Compliant / Needs Revision / Significant Issues] CLEAR VIOLATIONS [For each violation:] Issue: [description] Regulation: [specific cite] Problem: [what makes this non-compliant] Risk Level: [High/Medium/Low] Fix: [specific corrective action] GREY AREAS / INTERPRETIVE QUESTIONS [For each grey area:] Issue: [description] Regulation: [specific cite] Question: [what makes this ambiguous] Conservative Approach: [safest interpretation] Aggressive Approach: [more permissive interpretation] Recommendation: [which approach to take and why] COMPLIANT ELEMENTS [Brief confirmation of what is already compliant] REQUIRED ADDITIONS Missing disclosures or legends: - [What must be added] - [Where it must appear] - [Specific regulatory cite requiring it] RECORD-KEEPING REQUIREMENTS Documents that must be retained: - [Document type] - [Retention period] - [Regulatory cite] IMPLEMENTATION CHECKLIST [ ] All violations corrected [ ] Grey areas resolved with documented decision [ ] Required disclosures added [ ] Legal review completed [ ] Compliance officer approval obtained [ ] Record-keeping system updated ITEMS REQUIRING FURTHER RESEARCH [List any regulatory questions requiring consultation with specialized securities law counsel] ``` ### Drafting and Client Communication AI excels at drafting when given sufficient context and clear structure requirements. These prompts produce first drafts that require attorney review and refinement. #### Discovery Requests (Interrogatories and RFPs) **When to Use**: When drafting initial discovery requests or supplementing existing discovery. **Key Technique**: Tie each request directly to your case theory and make them specific enough to be non-objectionable. **Example Prompt:** ``` **INSTRUCTIONS** You are a litigation associate drafting discovery requests. Your requests should be specific, targeted, and difficult to object to on grounds of vagueness or overbreadth. Follow these guidelines: - Each request should seek specific, identifiable information - Define ambiguous terms in the definitions section - Use time limitations to narrow scope - Avoid compound requests (ask one thing at a time) - Include clear instructions for how to respond **CONTEXT** This is a breach of contract case. We represent the plaintiff, a software development company that contracted with defendant to build a custom inventory management system. Contract was executed: March 15, 2023 Performance due: September 15, 2023 Defendant delivered: November 30, 2023 (10 weeks late) System defects: Multiple critical bugs, incomplete features Our damages: - Lost profits from inability to track inventory ($150,000) - Cost of hiring another developer to fix defects ($75,000) - Reputational harm from poor inventory management ($50,000) Defendant's defenses: - Claims delays were due to plaintiff providing incomplete specifications - Claims "defects" are actually requested features plaintiff later rejected - Disputes causation of lost profits **INPUT** Draft 15 interrogatories and 10 requests for production of documents. Interrogatories should address: 1. The development timeline and reasons for delay 2. Communications about specifications and requirements 3. Knowledge of defects and attempts to fix them 4. The contract interpretation and performance expectations 5. Defendant's calculation of damages (if any counterclaim) Requests for Production should seek: 1. All versions of the software and code 2. Project management documents and timelines 3. Communications about specifications and requirements 4. Testing records and bug reports 5. Financial records showing costs incurred **OUTPUT** Format as formal discovery requests: [CAPTION] PLAINTIFF'S FIRST SET OF INTERROGATORIES TO DEFENDANT INSTRUCTIONS [Standard interrogatory instructions] DEFINITIONS [Define key terms that appear in multiple interrogatories] INTERROGATORIES INTERROGATORY NO. 1: Identify each person who participated in the initial contract negotiations in March 2023, including: (a) Full name and title (b) Dates of involvement (c) Role in negotiations (d) Whether they are still employed by Defendant [Continue through Interrogatory 15...] --- PLAINTIFF'S FIRST REQUEST FOR PRODUCTION OF DOCUMENTS TO DEFENDANT INSTRUCTIONS [Standard RFP instructions] DEFINITIONS [Define key terms] REQUESTS FOR PRODUCTION REQUEST NO. 1: All versions of the inventory management system software delivered to Plaintiff, including: (a) Source code (b) Compiled applications (c) Documentation provided with each version (d) Version control logs showing changes between versions Produce in native electronic format with metadata intact. [Continue through Request 10...] ATTESTATION I declare under penalty of perjury that the foregoing responses are true and correct. Date: ________________ _______________________________ [Attorney Name] Attorney for Plaintiff ``` #### Demand Letters **When to Use**: When initiating settlement discussions or making pre-litigation demands. **Key Technique**: Balance persuasive advocacy with professional tone; include specific demand and deadline. **Example Prompt:** ``` **INSTRUCTIONS** You are a litigation attorney drafting a pre-litigation demand letter. The letter should be: - Professional but firm in tone - Factually accurate and specific - Legally sound (cite relevant law) - Persuasive about liability and damages - Clear about the demand and deadline - Written to preserve the option of litigation Avoid: - Inflammatory language - Unsupported legal conclusions - Threats that cannot be carried out - Overly technical jargon **CONTEXT** We represent a homeowner (Maria Rodriguez) in a construction defect matter. She hired defendant contractor (BuildRight Construction) to remodel her kitchen for $85,000 (contract dated January 10, 2024). Problems: - Cabinets installed crooked (visibly uneven, doors don't close) - Countertop has visible seam in wrong location (not per plan) - Plumbing leaks discovered after completion (water damage to floor) - Electrical outlets not up to code (failed inspection) - Completion date missed by 6 weeks Financial impact: - Cost to repair defects: $32,000 (estimate from licensed contractor) - Water damage repair: $8,500 - Temporary housing during repairs: $6,000 - Diminution in home value: $15,000 Client wants to: 1. Get all work corrected to contract standards 2. Recover repair costs and consequential damages 3. Avoid litigation if possible (home equity line at stake) Jurisdiction: Texas (applicable law: Texas Residential Construction Commission Act, Texas Deceptive Trade Practices Act) **INPUT** Draft a demand letter to BuildRight Construction. The letter should: 1. Establish the attorney-client relationship 2. Summarize the contract and its terms 3. Detail each defect with specificity 4. Explain why these are breaches of contract 5. Reference relevant Texas law 6. Itemize damages with supporting evidence 7. Make a clear demand (correct defects OR pay damages) 8. Set a reasonable deadline (30 days) 9. Indicate willingness to discuss resolution 10. Preserve litigation option without explicit threats **OUTPUT** Format as a formal business letter: [Law Firm Letterhead] [Date] BuildRight Construction [Address] Re: Demand for Correction of Construction Defects Property: 123 Main Street, Austin, Texas Client: Maria Rodriguez Contract Date: January 10, 2024 Dear Mr. [Name]: This firm represents Maria Rodriguez concerning serious construction defects in the kitchen remodeling project you performed at her residence located at 123 Main Street, Austin, Texas. This letter serves as formal notice of breach of contract and demand for correction or compensation. [BACKGROUND section - contract details, timeline, scope of work] [DEFECTS section - detailed description of each defect with specific observations] [LEGAL BASIS section - breach of contract, DTPA violations, applicable law] [DAMAGES section - itemized with supporting documentation] [DEMAND section - specific demand, two options (fix or pay), deadline] [CONCLUSION - willingness to discuss, preservation of rights] Please respond to this demand by [DATE, 30 days from letter date]. We remain open to discussing a reasonable resolution that avoids the expense and uncertainty of litigation. However, if we do not receive a satisfactory response by the deadline, Ms. Rodriguez has authorized us to pursue all available legal remedies. Sincerely, [Attorney Name] [Attorney for Maria Rodriguez] Enclosures: - Contract dated January 10, 2024 - Repair estimate from [Contractor Name] - Water damage estimate - Photographs of defects - Failed inspection report ``` #### Client Status Updates and Case Explanations **When to Use**: When explaining legal developments to clients in plain language. **Key Technique**: Translate legal concepts without being condescending; always include next steps and timeline. **Example Prompt:** ``` **INSTRUCTIONS** You are a paralegal drafting a client communication. Write in plain English suitable for a client with no legal background (8th grade reading level). Guidelines: - Avoid legal jargon; when unavoidable, define terms in parentheses - Use active voice and short sentences - Focus on practical implications for the client - Be direct about risks and uncertainties - Always end with clear next steps and timeline Tone should be: - Professional but warm - Informative without being overwhelming - Honest about challenges without being alarmist - Reassuring without making promises **CONTEXT** We represent plaintiff in an employment discrimination case. Client is a 45-year-old woman (Jennifer Thompson) who was terminated from her job as a regional sales manager. She alleges age and gender discrimination. Recent development: The defendant filed a Motion for Summary Judgment arguing that: 1. Plaintiff cannot establish a prima facie case of discrimination 2. Defendant had legitimate, non-discriminatory reasons for termination (poor performance) 3. Plaintiff cannot show these reasons were pretextual We have 30 days to respond. We believe we have strong evidence: - Performance reviews were consistently "exceeds expectations" until new manager arrived - Statistical evidence showing women over 40 are terminated at higher rates - Email from supervisor calling her "past her prime" - Younger, less qualified male was promoted to her position However, there are some weaknesses: - Last performance review (just before termination) was negative - Client did miss some sales targets in final quarter - Email comment is arguably ambiguous **INPUT** Draft an email to Jennifer Thompson explaining: 1. What a Motion for Summary Judgment is and why it matters 2. What the defendant is arguing 3. What we need to do to oppose it 4. Our assessment of strengths and weaknesses 5. What happens next and when 6. What we need from her (if anything) Keep it to 2-3 pages maximum. Use section headings for readability. **OUTPUT** Subject: Important Update: Defendant's Motion for Summary Judgment Dear Jennifer, I'm writing to update you on an important development in your case. The defendant has filed what's called a "Motion for Summary Judgment," and I want to explain what this means, what we're doing about it, and what comes next. **What Is a Motion for Summary Judgment?** [Explanation in plain language - what it is, what it seeks, why defendants file these motions] **What Is the Defendant Arguing?** [Summary of the three main arguments in accessible language] **Our Response Strategy** [Explanation of our opposing arguments and the evidence we'll use] **Honest Assessment of Where We Stand** The good news: [List of strengths in plain language] The challenges: [List of weaknesses, explained honestly but not alarmingly] Bottom line: [Realistic assessment of likely outcome] **What Happens Next** Here's the timeline: - [Date]: Our response is due - [Date range]: We'll prepare your declaration and gather additional evidence - [Date]: Judge will hear arguments (you don't need to attend) - [Date range]: Judge will issue a ruling **What We Need From You** [Specific action items with deadlines, if any] **What This Means for Your Case** [Explanation of the two possible outcomes: 1. If we win the motion 2. If we lose the motion] **Questions?** I know this is a lot to digest. Please don't hesitate to call me at [phone] or email if you have any questions. I'm here to help you understand what's happening at every stage. We believe in your case and are working hard to achieve the best possible outcome. Best regards, [Paralegal Name] [Law Firm Name] [Contact Information] ``` #### Negotiation Strategy and Talking Points **When to Use**: When preparing for settlement discussions, mediation, or negotiation meetings. **Key Technique**: Request both your arguments and anticipated counterarguments to prepare for back-and-forth. **Example Prompt:** ``` **INSTRUCTIONS** You are a mediator and negotiation strategist helping prepare for a settlement conference. Provide objective analysis of both parties' positions, including: - Strong arguments on each side - Weaknesses in each position - Likely negotiation ranges - Strategic approaches for different scenarios Be realistic about both strengths and weaknesses. Do not inflate the value of weak arguments. **CONTEXT** Commercial lease dispute. We represent the landlord (Thompson Properties LLC). Background: - 10-year retail lease executed in 2019 - Tenant (Ace Sporting Goods) stopped paying rent in March 2024 - Tenant claims: Store became economically unviable due to road construction blocking access for 8 months; lease should be terminated or reduced due to "frustration of purpose" and "constructive eviction" - Landlord position: No lease provision excuses rent for road construction; tenant assumed this risk; other tenants in same plaza continued paying Financials: - Monthly rent: $15,000 - Unpaid rent to date: $120,000 (8 months) - Remaining lease term: 5 years - Total exposure if tenant walks: $900,000 plus re-leasing costs - Tenant's proposed settlement: $30,000 to terminate lease - Our costs to re-lease: ~$150,000 (tenant improvements, broker fees, lost rent during vacancy) Market context: - Similar retail space is in low demand due to e-commerce shift - Road construction is now complete - Finding new tenant could take 12-18 months - New tenant would likely demand lower rent ($10,000-12,000/month) Mediation is scheduled for next week. We have authority to settle for as low as $200,000 cash payment plus immediate possession. **INPUT** Prepare comprehensive negotiation strategy including: 1. Our strongest arguments for full enforcement 2. Tenant's strongest arguments for release 3. Weaknesses in our position we must address 4. Realistic assessment of litigation outcomes if we don't settle 5. Negotiation ranges and likely settlement zone 6. Opening position, target settlement, walk-away point 7. Tactical approaches for different negotiation scenarios 8. Responses to anticipated tenant arguments 9. Creative deal structures beyond just cash payment **OUTPUT** Format as negotiation preparation memo: NEGOTIATION STRATEGY MEMORANDUM CLIENT: Thompson Properties LLC MATTER: Ace Sporting Goods Lease Dispute MEDIATION DATE: [Date] I. EXECUTIVE SUMMARY [Two paragraphs: current situation and recommended settlement approach] II. OUR STRONGEST ARGUMENTS A. Legal Arguments 1. [Argument with supporting law] 2. [Argument with supporting law] 3. [Argument with supporting law] B. Practical/Business Arguments 1. [Practical point] 2. [Practical point] C. Credibility/Equitable Arguments [Why our position is fair/reasonable] III. ANTICIPATED TENANT ARGUMENTS A. Tenant's Legal Arguments 1. [Their argument] - Our response: [how we counter this] 2. [Their argument] - Our response: [how we counter this] B. Tenant's Practical Arguments 1. [Their practical point] - Our response: [our counter] IV. HONEST ASSESSMENT OF WEAKNESSES IN OUR POSITION 1. [Weakness] - Significance: [how much this matters] - Mitigation strategy: [how we address it in negotiation] 2. [Weakness] - Significance: [how much this matters] - Mitigation strategy: [how we address it in negotiation] V. LITIGATION RISK ANALYSIS If we proceed to trial: Best case scenario: - Outcome: [what we win] - Probability: [realistic percentage] - Recovery: [amount and timing] - Costs: [litigation costs] - Net result: [bottom line] Worst case scenario: - Outcome: [what we lose] - Probability: [realistic percentage] - Exposure: [potential loss] - Costs: [litigation costs] - Net result: [bottom line] Most likely scenario: - Outcome: [realistic expectation] - Probability: [realistic percentage] - Recovery: [amount and timing] - Net result: [bottom line] VI. SETTLEMENT ZONE ANALYSIS Tenant's best alternative to settlement: $0 (wins on frustration doctrine) Tenant's worst alternative: $900,000 (loses completely) Tenant's likely range: $0 - $100,000 Our best alternative: $900,000 (tenant pays everything) Our worst alternative: $0 (tenant wins release) plus $150,000 re-leasing costs = -$150,000 Our likely range: $200,000 - $400,000 ZONE OF POSSIBLE AGREEMENT (ZOPA): $100,000 - $400,000 VII. NEGOTIATION POSITIONS Opening Position: $500,000 cash payment plus immediate possession - Justification: [reasoning we'll articulate] Target Settlement: $300,000 cash payment plus immediate possession, or $250,000 plus tenant completes some repairs - Justification: [why this is fair to both sides] Walk-Away Point: $200,000 cash payment plus immediate possession - Rationale: [why we won't go lower] VIII. TACTICAL APPROACHES A. If Tenant Opens Low ($50,000 or less): - Response: [how we react] - Next move: [our counter] B. If Tenant Starts at Reasonable Number ($150,000-200,000): - Response: [how we react] - Next move: [our counter] C. If Tenant Refuses to Pay Anything: - Response: [how we react] - Next move: [whether we walk or try alternative structures] IX. CREATIVE DEAL STRUCTURES Beyond cash payment, consider: 1. Tenant Remains but Modified Terms - Reduce rent to $10,000 for remainder of term - Tenant pays $X towards arrears over time - Early termination option in Year 3 2. Hybrid Approach - Cash payment of $[X] - Tenant helps find replacement tenant (assigns lease) - Tenant leaves improvements in place 3. Release Plus Marketing Support - Cash payment of $[X] - Tenant displays "For Lease" signs in windows - Tenant allows showings during final 60 days X. LIKELY QUESTIONS FROM MEDIATOR Q: Why not just let them out and find a new tenant? A: [Our response] Q: Isn't this just throwing good money after bad? A: [Our response] Q: What if tenant files bankruptcy? A: [Our response] XI. KEY TALKING POINTS FOR MEDIATION When mediator is in our room: - [Point 1] - [Point 2] - [Point 3] When mediator is in their room (what we want mediator to tell them): - [Point 1] - [Point 2] - [Point 3] XII. AUTHORITY AND DECISION TREE We have authority to settle for $200,000 minimum. Decision tree: - If offer is $200,000+: Accept - If offer is $150,000-$199,000: Call client for additional authority - If offer is under $150,000: Reject and prepare for litigation XIII. POST-MEDIATION PLAN If we settle: - [Implementation steps] If we don't settle: - [Litigation next steps] - [Timeline] - [Budget] ``` ### Trial Preparation and Strategy Trial preparation requires meticulous organization and strategic analysis. These prompts help prepare for witness examination, organize evidence, and draft trial documents. #### Witness Preparation and Cross-Examination Planning **When to Use**: When preparing to examine a witness or preparing your own witness for testimony. **Key Technique**: Organize questions by topic, anticipate objections, and prepare for unexpected answers. **Example Prompt:** ``` **INSTRUCTIONS** You are a trial attorney preparing for witness examination. Develop a comprehensive examination outline that: - Follows logical topic progression - Uses open-ended questions for friendly witnesses - Uses leading questions for adverse witnesses - Anticipates objections and has responses ready - Identifies exhibit foundation requirements - Flags areas where witness may be evasive **CONTEXT** Personal injury trial: plaintiff injured in car accident. We represent plaintiff. This is direct examination of plaintiff's treating physician (Dr. Sarah Chen, orthopedic surgeon). She treated plaintiff for 18 months following the accident. Key testimony we need: 1. Establish her qualifications (expert witness) 2. Nature and extent of plaintiff's injuries 3. Treatment provided and medical necessity 4. Plaintiff's pain and limitations 5. Future medical needs and costs 6. Causation (injuries caused by accident, not pre-existing) 7. Permanency of injuries Challenges: - Defense will argue some injuries are pre-existing (plaintiff had prior back pain) - Defense may challenge treatment as excessive/unnecessary - Defense will argue plaintiff could return to work sooner - Medical bills total $287,000 (defense says only $150,000 was necessary) Exhibits to use: - Exhibit 12: Dr. Chen's CV - Exhibit 15-23: Medical records and imaging - Exhibit 24: Life care plan prepared by Dr. Chen - Exhibit 25: Fee invoice **INPUT** Prepare a direct examination outline for Dr. Chen. Include: 1. Qualification section (establish expertise) 2. Background on how she came to treat plaintiff 3. Initial examination and diagnosis 4. Treatment rendered over 18 months 5. Current condition and prognosis 6. Future medical needs and costs 7. Causation opinions 8. Cross-references to exhibits 9. Anticipated defense objections with responses 10. Backup questions if witness answer isn't clear **OUTPUT** DIRECT EXAMINATION OUTLINE: DR. SARAH CHEN I. INTRODUCTION AND QUALIFICATIONS [Exhibit 12] A. Education and Training Q: Please state your full name and occupation. A: [Expected answer] Q: Where did you attend medical school? A: [Expected] Q: Tell the jury about your specialized training in orthopedic surgery. A: [Expected - residency, fellowship] [Continue with qualification questions...] OFFER DR. CHEN AS EXPERT WITNESS IN ORTHOPEDIC SURGERY Anticipated objection: [None expected] B. Experience and Practice Q: How long have you been practicing orthopedic surgery? Q: Approximately how many patients have you treated for injuries similar to [Plaintiff]'s? [Continue...] II. INITIAL TREATMENT OF PLAINTIFF A. First Encounter Q: Dr. Chen, do you recognize the woman seated at counsel table in the blue dress? A: [Identifies plaintiff] Q: How did [Plaintiff] come to be your patient? A: [Referred by ER, date was [DATE]] Anticipated objection: Hearsay if she describes what patient told her Response: Not offered for truth; offered to explain basis of her diagnosis and treatment [Continue with initial examination questions...] B. Diagnosis [Reference Exhibits 15-16: X-rays and MRI] Q: After your initial examination, what did you diagnose? A: [Expected: lumbar spine fracture, soft tissue damage] Q: Let me show you what's been marked as Exhibit 15. Do you recognize this? A: [Yes, X-ray from initial exam] MOVE TO ADMIT EXHIBIT 15 Anticipated objection: Lack of foundation Response: Dr. Chen has testified she ordered this, reviewed it, and relied on it in treatment Q: Dr. Chen, what does this X-ray show? [Continue with imaging testimony...] III. TREATMENT PROVIDED A. Initial Treatment Phase (Months 1-6) Q: What treatment did you recommend initially? A: [Conservative treatment: PT, medication] Q: Why did you recommend this course of treatment? A: [Medical necessity explanation] Anticipated defense argument: Treatment was excessive Preparation: Have Dr. Chen explain standard of care, why each treatment was medically necessary [Continue with treatment timeline...] IV. CAUSATION Q: Dr. Chen, based on your examination and review of records, do you have an opinion to a reasonable degree of medical certainty whether the injuries you've described were caused by the accident on [DATE]? Anticipated objection: Assumes facts not in evidence, calls for speculation Response: Dr. Chen has testified to her review of records and examination; she can opine on causation based on this Q: What is that opinion? A: [Injuries caused by accident] Q: What is the basis for that opinion? A: [Timing, mechanism of injury consistent, no prior imaging showing these injuries] ADDRESSING PRE-EXISTING CONDITION: Q: Doctor, you're aware plaintiff had complained of occasional back pain before this accident? A: [Yes] Q: How does that affect your opinion? A: [Acute injury superimposed on chronic condition; clear aggravation; different pain pattern] [Continue with causation testimony...] V. PROGNOSIS AND FUTURE TREATMENT [Exhibit 24: Life Care Plan] Q: Dr. Chen, do you have an opinion about whether plaintiff's injuries are permanent? Q: What future medical treatment will plaintiff require? [List each item from life care plan with medical necessity explanation] Q: Have you prepared a life care plan documenting these needs? MOVE TO ADMIT EXHIBIT 24 Q: What is the estimated cost of this future medical care? A: [Approximately $180,000 over lifetime] Anticipated defense objection: Speculative, no foundation for costs Response: Dr. Chen can testify to medical necessity; costs are based on current rates (if challenged, we have economist expert) VI. IMPACT ON PLAINTIFF'S LIFE Q: Doctor, based on your treatment of plaintiff over 18 months, what functional limitations does she continue to experience? A: [Cannot lift, prolonged sitting/standing limited, chronic pain] Q: Can she return to her previous work as a warehouse supervisor? A: [No, requires physical activity beyond her restrictions] VII. FEES AND BILLING [Exhibit 25] Q: What were your fees for treating plaintiff? A: [Total from bill] Q: Were these fees reasonable and customary for your area? A: [Yes, standard rates] Q: Were all treatments medically necessary? A: [Yes] Anticipated objection: Self-serving Response: She's qualified to testify about standard fees and medical necessity TENDER THE WITNESS --- POTENTIAL PROBLEM AREAS: If defense asks on cross about: 1. Pre-existing condition: Dr. Chen should emphasize acute vs. chronic; mechanism of injury; timeline 2. Gaps in treatment: Explain COVID, insurance issues, patient compliance (but don't blame patient) 3. Overly aggressive treatment: Dr. Chen should cite standard of care, peer-reviewed literature 4. Ability to return to work: Emphasize specific functional limitations that preclude her job duties REDIRECT PREPARATION: Have ready: [list of clarifying questions based on likely cross themes] ``` #### Exhibit Organization and Cross-Reference System **When to Use**: When organizing trial exhibits and creating exhibit lists for trial. **Key Technique**: Create a system that allows quick retrieval during trial and shows relationships between exhibits. **Example Prompt:** ``` **INSTRUCTIONS** You are a trial paralegal creating a comprehensive exhibit management system. The goal is to enable the trial team to quickly locate exhibits during testimony and understand how exhibits relate to each other. Create clear categorization, cross-references, and usage notes for each exhibit. **CONTEXT** Construction defect trial. We represent homeowners. Trial is 3 weeks away. We have 87 exhibits covering: - Contract documents (original agreement, change orders) - Correspondence (emails, letters between parties) - Photographs (defects, work in progress, completed work) - Expert reports (our expert and defense expert) - Financial documents (invoices, payment records, repair estimates) - Technical documents (building codes, inspection reports, warranties) Witnesses who will use exhibits: - Homeowner (client) - Contractor (defendant) - Our construction expert - Defense construction expert - Building inspector - Repair contractor who gave estimates **INPUT** Analyze the attached 87 exhibits and create: 1. Master exhibit list with complete information for each 2. Categorization by type and topic 3. Witness-by-witness exhibit usage plan 4. Cross-reference system showing which exhibits reference others 5. Chronological event timeline with exhibit references 6. Quick-reference guide for trial team For each exhibit, identify: - Exhibit number - Document type - Date - Author/source - Brief description (1-2 sentences) - Which witness(es) will use it - What it proves (relevance) - Related exhibits (cross-references) - Foundation requirements - Potential objections **OUTPUT** EXHIBIT MANAGEMENT SYSTEM I. MASTER EXHIBIT LIST CONTRACTS AND AGREEMENTS Exhibit 1: Original Construction Contract - Date: March 15, 2023 - Parties: [Homeowner] and [Contractor] - Description: 12-page contract for home addition, price $285,000 - Witnesses: Homeowner, Contractor - Proves: Scope of work, price, timeline, warranty obligations - Related exhibits: 2 (Change Order #1), 3 (Change Order #2) - Foundation: Homeowner will identify and authenticate - Potential objections: None anticipated [Continue for all 87 exhibits...] --- II. EXHIBITS BY CATEGORY A. CONTRACT DOCUMENTS - Exhibit 1: Original Contract - Exhibit 2: Change Order #1 (added bathroom, +$35,000) - Exhibit 3: Change Order #2 (upgraded countertops, +$8,000) - Exhibit 4: Final payment invoice B. CORRESPONDENCE - DEFECT COMPLAINTS - Exhibit 10: Email from homeowner to contractor, 8/15/23 (first complaint) - Exhibit 11: Contractor response, 8/18/23 - Exhibit 12: Email chain, 9/1-9/5/23 (escalating complaints) [Continue...] C. PHOTOGRAPHS - DEFECTS - Exhibits 20-35: Foundation cracks - Exhibits 36-42: Roof leaks - Exhibits 43-48: Electrical issues [Continue...] --- III. WITNESS-BY-WITNESS EXHIBIT PLAN WITNESS 1: [HOMEOWNER NAME] (Plaintiff) Opening Statement Exhibits: - Exhibit 1 (Contract - establish agreement) - Exhibit 45 (Photo of dream home before construction) - Exhibit 73 (Photo of home with visible defects) Direct Examination Flow: Topic 1: Entering the Contract - Exhibit 1: Contract (authenticate, discuss terms) - Exhibit 5: Marketing materials from contractor Topic 2: Work Performance Issues - Exhibit 10: First complaint email - Exhibit 12: Email chain - Timeline: Show Exhibits 10, 12, 15, 18 chronologically Topic 3: Discovery of Defects - Exhibits 20-35: Foundation photos (select best 5 to show) Priority: Ex. 20, 23, 28, 32, 35 - Exhibits 36-42: Roof photos (select best 3) Priority: Ex. 36, 39, 42 Topic 4: Attempts to Get Repairs - Exhibit 16: Demand letter - Exhibit 17: Contractor's refusal Topic 5: Financial Impact - Exhibit 4: What we paid - Exhibit 75: Repair estimate - Exhibit 78: Lost value appraisal --- WITNESS 2: [CONTRACTOR NAME] (Defendant) Cross-Examination Strategy: Topic 1: Contract Obligations - Exhibit 1: Contract (confront with specific warranty provisions) - Exhibit 4: Invoice showing full payment received Topic 2: Knowledge of Defects - Exhibit 10: Homeowner's first complaint (he received it) - Exhibit 11: His inadequate response - Exhibit 12: Ignored multiple follow-up complaints Topic 3: Departures from Contract - Exhibit 6: Contract specs for materials - Exhibit 50: Photo of inferior materials actually used --- WITNESS 3: [OUR EXPERT NAME] Direct Examination: Topic 1: Qualification - Exhibit 60: Expert's CV Topic 2: Inspection and Findings - Exhibit 61: Expert report - Exhibits 20-48: Photos (expert will explain each defect) - Exhibit 62: Building code sections violated Topic 3: Causation and Repair Costs - Exhibit 75: Repair cost estimate (expert prepared this) - Exhibit 76: Methodology for estimating costs --- IV. CHRONOLOGICAL EXHIBIT TIMELINE March 15, 2023: Contract signed - Exhibit 1: Contract April 1, 2023: Work begins - Exhibit 7: Start date acknowledgment May 20, 2023: Change Order #1 - Exhibit 2: Change order [Continue chronologically through all key dates...] --- V. CROSS-REFERENCE MAP Exhibit 1 (Contract) is referenced in: - Exhibit 2 (Change Order #1 - references original contract) - Exhibit 4 (Final invoice - references contract) - Exhibit 10 (Complaint email - references warranty provisions) - Exhibit 16 (Demand letter - cites contract breaches) - Exhibit 61 (Expert report - compares work to contract specs) Exhibit 20 (Foundation photo #1) is referenced in: - Exhibit 61 (Expert report - discussed on page 4) - Exhibit 75 (Repair estimate - foundation repairs) [Continue for all cross-referenced exhibits...] --- VI. FOUNDATION REQUIREMENTS Exhibits 1-5 (Contract documents): - Foundation: Homeowner testimony + business records exception - Witness: Homeowner identifies and authenticates - Objections: None anticipated Exhibits 20-48 (Photographs): - Foundation: Photographer testimony OR homeowner if she took them - Witness: Verify accurate representation of what it shows - Check EXIF data: Does it show date taken? - Potential objection: Not accurate representation Exhibit 61 (Expert report): - Foundation: Expert testimony - Pre-trial: Confirm expert report was timely disclosed - Potential objection: Late disclosure (verify disclosure date) --- VII. QUICK-REFERENCE GUIDE FOR TRIAL TEAM BEST EXHIBITS TO SHOW JURY (highest impact): 1. Exhibit 1: Contract (what we agreed to) 2. Exhibit 20: Foundation crack photo (most dramatic) 3. Exhibit 39: Roof leak photo (water damage visible) 4. Exhibit 50: Side-by-side of spec'd vs. actual materials 5. Exhibit 75: Repair cost estimate ($127,000) "SMOKING GUN" EXHIBITS: - Exhibit 12: Email where contractor admits "corners were cut" - Exhibit 50: Photo proving inferior materials used EXHIBITS TO ENLARGE FOR JURY: - Exhibits 20, 23, 28, 39, 42, 50 (key photos) - Exhibit 1, page 4 (warranty provision) EXHIBIT BINDERS: - Judge's binder: All 87 exhibits, tabbed - Jury binder: 25 key exhibits only - Witness binders: Only exhibits each witness will use TECHNOLOGY NEEDS: - Load Exhibits 20-48 into presentation software for quick display - Create video deposition clips synced with exhibits - Have backup paper copies of all technology-displayed exhibits EXHIBIT TRACKING DURING TRIAL: □ Exhibit offered □ Objection (if any) □ Ruling □ Admitted □ Published to jury □ Used with witness ``` #### Motion in Limine Strategy **When to Use**: When preparing pre-trial motions to exclude or admit evidence. **Key Technique**: Identify specific evidence to exclude, legal basis, and practical trial impact. **Example Prompt:** ``` **INSTRUCTIONS** You are a trial attorney developing motions in limine strategy. For each potential motion, provide: - The specific evidence to be excluded - Legal basis with case citations - Practical impact on trial if granted - Likelihood of success - Strategic considerations Be realistic about which motions are likely to succeed. Don't waste time on motions courts routinely deny. **CONTEXT** Medical malpractice trial. We represent plaintiff (patient). Trial is 6 weeks away. Background: - Plaintiff underwent knee replacement surgery - Claims surgeon damaged nerve causing permanent foot drop - Defense: Nerve damage is a known risk; no negligence occurred Evidence concerns we have: 1. Defense wants to introduce plaintiff's prior medical history including mental health treatment for depression (irrelevant and prejudicial) 2. Defense retained expert who will testify surgery met standard of care (we have questions about his qualifications - only did 5 knee surgeries in last 5 years) 3. Defendant surgeon posted on social media about "frivolous lawsuits" after this case was filed (we want to use; defense will try to exclude) 4. Plaintiff was previously convicted of misdemeanor shoplifting 10 years ago (defense wants to impeach her with this; we want to exclude) 5. Hospital's internal incident report documenting the nerve damage (defense claims attorney-client privilege; we want to use it) 6. Defense wants to argue plaintiff's damages are minimal because she can use a wheelchair (we want to preclude this argument as improper) **INPUT** Identify and prioritize motions in limine we should file. For each motion, provide: 1. Motion title and what we're seeking 2. Specific evidence to be excluded/admitted 3. Legal standard and governing law 4. Supporting arguments with case citations (if you know them; otherwise note "research required") 5. Anticipated opposition arguments 6. Our response to opposition 7. Likely ruling (granted/denied/granted in part) 8. Strategic importance (high/medium/low) 9. Whether to brief this or handle at hearing **OUTPUT** MOTIONS IN LIMINE STRATEGY RECOMMENDED MOTIONS (in priority order) --- MOTION #1: Exclude Plaintiff's Mental Health History SEEKING: Order precluding defendant from introducing evidence of plaintiff's treatment for depression. SPECIFIC EVIDENCE TO EXCLUDE: - Plaintiff's medical records from Dr. Smith (psychiatrist), 2020-2022 - Any testimony about plaintiff seeking mental health treatment - Any references to depression, anxiety, or psychiatric medication LEGAL STANDARD: Federal Rule of Evidence 401-403 analysis: - FRE 401: Evidence is relevant if it has tendency to make material fact more/less probable - FRE 403: Court may exclude relevant evidence if probative value is substantially outweighed by danger of unfair prejudice Medical privacy: [State] law limits disclosure of mental health records; strong public policy favoring confidentiality ARGUMENTS IN SUPPORT: 1. Not Relevant (FRE 401) - Mental health history has no bearing on whether surgeon was negligent - Depression does not affect credibility about physical injury - No causal connection between mental health and physical injury 2. Unfair Prejudice Substantially Outweighs Any Probative Value (FRE 403) - Jury may improperly discount plaintiff's testimony due to mental health stigma - Risk of jury speculation about mental health issues - Will confuse issues and waste time with collateral matter 3. Protected Health Information - [State] mental health privilege statute - Strong public policy protecting mental health treatment confidentiality ANTICIPATED DEFENSE ARGUMENTS: - "Goes to damages" - depression contributed to plaintiff's suffering, we can't separate physical from mental - "Goes to credibility" - mental health affects perception and testimony - "Plaintiff opened the door" - claiming emotional distress damages OUR RESPONSE: - Damages for foot drop are objectively verifiable (medical records, physical examination, functional limitations) - No claim for emotional distress damages separate from physical injury - Standard for relevance not met; any marginal relevance destroyed by prejudice CASE LAW TO RESEARCH: - [Note: Research cases in this jurisdiction on admissibility of mental health history in physical injury cases] - Similar motions granted in: [research required] LIKELY RULING: **GRANTED** - Courts routinely exclude mental health history in physical injury cases absent emotional distress claim - Strong FRE 403 argument STRATEGIC IMPORTANCE: **HIGH** - Critical to prevent jury bias - Plaintiff will testify; cannot have jury discounting her credibility BRIEFING RECOMMENDATION: **Full written motion with brief** - Important issue requiring detailed legal analysis - Need to preserve for appeal if denied --- MOTION #2: Exclude Reference to Plaintiff's Prior Criminal Conviction SEEKING: Order precluding defendant from impeaching plaintiff with 10-year-old misdemeanor shoplifting conviction. SPECIFIC EVIDENCE TO EXCLUDE: - Any mention of 2014 shoplifting conviction - Any questions about criminal history - Any suggestion plaintiff is dishonest or untrustworthy LEGAL STANDARD: Federal Rule of Evidence 609: Impeachment by evidence of criminal conviction FRE 609(b): Conviction more than 10 years old is admissible only if: (1) Probative value substantially outweighs prejudicial effect, AND (2) Proponent gives adverse party reasonable written notice FRE 609(a)(2): Crimes involving dishonesty automatically admissible IF within 10 years Key question: When did conviction occur vs. when was sentence completed? - If conviction date controls: More than 10 years, FRE 609(b) applies - If release date controls: May be within 10 years ARGUMENTS IN SUPPORT: 1. More Than 10 Years Old - Conviction date: 2014 (11 years ago) - Even if sentence completion date used, was 2015 (10 years ago) - FRE 609(b) presumption against admission 2. Probative Value Does Not Substantially Outweigh Prejudice - Remote in time - not probative of current credibility - Minor offense - not reflective of character - Plaintiff has lived law-abiding life for decade since - Prejudicial effect: Jury will view plaintiff as dishonest, criminal - Unfairly bolsters defense: "She's a thief, don't trust her" 3. Defendant Has Not Given Required Notice - FRE 609(b) requires reasonable written notice - Defendant has not provided written notice of intent to use conviction - Prejudice to plaintiff in preparing to address this ANTICIPATED DEFENSE ARGUMENTS: - "Crime of dishonesty" under FRE 609(a)(2) - automatically admissible - Goes to credibility - theft shows dishonesty - Not overly prejudicial in civil case OUR RESPONSE: - Even crimes of dishonesty must meet 609(b) requirements when remote - 609(b) creates presumption against admission for old convictions - Shoplifting over decade ago says nothing about credibility today - In civil case, lower standard for unfair prejudice CASE LAW TO RESEARCH: - [Research: Circuit split on conviction date vs. release date for 10-year calculation] - [Research: Cases in this jurisdiction on admitting remote shoplifting convictions] LIKELY RULING: **GRANTED** - Strong argument under FRE 609(b) - 10-year-old misdemeanor unlikely to be admitted - Even if denied, court will likely give limiting instruction STRATEGIC IMPORTANCE: **HIGH** - Plaintiff must testify - Cannot have jury thinking "she's a thief" - May affect damages even if liability established BRIEFING RECOMMENDATION: **Full written motion** - Complex FRE 609 analysis - Need to fully develop arguments --- MOTION #3: Admit Defendant's Social Media Posts SEEKING: Order permitting plaintiff to introduce defendant surgeon's social media posts about frivolous lawsuits. [Continue with same detailed format...] --- MOTION #4: Exclude Defense Expert [Name] SEEKING: Order excluding defense expert on grounds of inadequate qualifications. [Continue with same detailed format...] --- MOTION #5: Admit Hospital Incident Report SEEKING: Order compelling production of and admitting hospital's internal incident report. [Continue with same detailed format...] --- MOTION #6: Preclude "Wheelchair" Argument SEEKING: Order precluding defense from arguing damages are minimal because plaintiff can use wheelchair. [Continue with same detailed format...] --- STRATEGIC PRIORITIES: MUST FILE (high importance, likely to win): 1. Motion #1 (mental health) 2. Motion #2 (prior conviction) SHOULD FILE (important issues): 3. Motion #4 (defense expert) 5. Motion #5 (incident report) CONSIDER FILING (tactical advantages): 3. Motion #3 (social media) 6. Motion #6 (wheelchair argument) TIMING: - File Motions #1, #2, #4, #5 together: [DATE, 4 weeks before trial] - Reserve Motion #3, #6 for hearing if strategic advantage emerges HEARING STRATEGY: - Lead with strongest motion (#1) - Be prepared to address all motions in one hearing - Have settlement leverage if we win key motions ``` ### Chapter Summary This chapter provided comprehensive, ready-to-use prompt templates organized by legal task: **Discovery and Document Review**: Deposition summaries, privilege logs, exhibit tagging, TAR training **Legal Research and Analysis**: Multi-jurisdictional statutory comparison, case law synthesis, regulatory compliance analysis **Drafting and Client Communication**: Discovery requests, demand letters, client updates, negotiation strategy **Trial Preparation and Strategy**: Witness examination outlines, exhibit management systems, motions in limine All examples follow the C.A.S.E. Framework and can be customized for your specific matters. Remember: * Always verify AI-generated legal conclusions * Redact confidential information before using public AI tools * Treat all AI output as a first draft requiring attorney review * Build your own library of successful prompts In the next chapter, we'll address the critical ethical considerations that must guide all AI use in legal practice, including verification duties, confidentiality protection, and professional responsibility compliance. *** *In Chapter 4, we'll explore the ethical guardrails and mandatory review protocols that ensure your AI use complies with the Rules of Professional Conduct.* # 4. Ethical Guardrails and Professional Responsibility Successful AI integration relies not just on technical skill but on adherence to professional standards. This chapter defines the ethical boundaries and mandatory verification steps necessary to safeguard client interests, maintain confidentiality, and comply with the Rules of Professional Conduct. **ATTENTION: THE DUTY OF CONFIDENTIALITY AND COMPETENCE REMAINS WITH THE LEGAL PROFESSIONAL.** AI is a powerful tool, but like any tool, it must be used responsibly and competently. Your professional license—and your client's interests—depend on understanding and following these ethical requirements. ### The Non-Delegable Duty of Verification Model Rule of Professional Conduct 1.1 (Competence) requires lawyers to understand the benefits and risks of new technology. The primary risk of generative AI is **"hallucination"**—the production of non-existent facts or legal precedent. The attorney and paralegal are solely responsible for all work product submitted to a court or client. #### The "Final Eye" Principle AI output is a starting point, never an endpoint. Every piece of work must be reviewed and adopted by a human professional. **Rule 11 Compliance**: Filing a document with invented case law (a hallucination) violates Federal Rule of Civil Procedure 11 (or equivalent state rules), which requires that all submissions have legal and factual basis. Sanctions, including financial penalties and public reprimand, are possible. **Real-World Example**: In *Mata v. Avianca, Inc.*, 2023 WL 4114965 (S.D.N.Y. June 22, 2023), attorney Steven Schwartz submitted a brief containing six non-existent cases generated by ChatGPT. The court sanctioned both the attorney and his law firm, stating: "Technological advances are commonplace and there is nothing inherently improper about using a reliable artificial intelligence tool for assistance... But existing rules impose a gatekeeping role on attorneys to ensure the accuracy of their filings." The court continued: "The Court is presented with an unprecedented circumstance... Six of the submitted cases appear to be bogus judicial decisions with bogus quotes and bogus internal citations." #### Verification Requirements for Legal Professionals **Paralegal Example**: An AI-assisted draft of a motion cites *Johnson v. Acme Corp., 12 F.3d 456 (9th Cir. 2021)*. The paralegal's mandatory step is to **independently search Westlaw or Lexis** for this exact citation to confirm: 1. The case exists 2. The holding supports the proposition cited 3. It is still **Good Law** (has not been overruled) 4. The quotations are accurate 5. The procedural posture is correctly stated **Lawyer Example**: A lawyer uses AI to summarize key facts from 20 exhibits. Before filing the Statement of Facts, the lawyer must **trace every single factual assertion** in the AI's summary back to the specific page and line number of the corresponding source exhibit or transcript. #### The Three-Layer Verification System Implement this three-layer approach for all AI-generated legal work: **Layer 1: AI Self-Review** Ask the AI to review its own work for potential errors or uncertainties: ``` Review your previous response and identify: 1. Any legal conclusions you are less than highly confident about 2. Citations that should be independently verified 3. Areas where the law may be unsettled or evolving 4. Any assumptions you made due to incomplete information ``` **Layer 2: Cross-Model Verification** For critical conclusions, use a different AI model to verify the first model's output: ``` Review the legal analysis below [paste first AI's response]. Identify any potential errors, omissions, or areas requiring additional research. Focus particularly on: - Citation accuracy - Logical consistency - Completeness of analysis - Conflicting authorities that may exist ``` **Layer 3: Human Verification** The attorney or supervised paralegal must: * Verify every citation in primary sources (Westlaw, Lexis, or official reporters) * Confirm factual assertions against source documents * Exercise professional judgment on legal conclusions * Ensure the analysis addresses the specific client situation * Add appropriate qualifications and disclaimers ### Protecting Client Confidentiality The duty of confidentiality (Model Rule 1.6) is paramount. AI use must be consistent with the obligation to protect client information. #### Public vs. Secure Environments **Public Models (e.g., general web chatbots)**: **NEVER** use these platforms for any text containing: * Client names or case names * Unredacted transcripts or documents * Addresses or specific identifying information * Specific facts not already publicly available * Privileged communications * Attorney work product * Confidential client information Assume anything you type into a public model may be used to train that model or could potentially be accessed by others, thus breaching confidentiality. **Secure Models (e.g., firm-approved, closed-environment tools)** These platforms are built with specific privacy agreements that assure client data remains secure and is not used for external model training. Only use these tools for sensitive data when: * Your firm has vetted and approved the platform * A Business Associate Agreement (BAA) or similar contract is in place * The platform explicitly commits to not using your data for training * Adequate security measures are documented * The platform complies with relevant data protection regulations #### Stripping Identifying Information For low-stakes, non-sensitive tasks that require a general AI tool (e.g., drafting an email template), practice robust redaction. **Paralegal Example**: The paralegal wants AI to help clarify a confusing client email about a timeline. The paralegal **replaces all proper nouns** (names, company names, cities, specific product names) with generic placeholders before submitting the query: **Before (Confidential - DO NOT USE):** ``` Sarah Martinez from TechCorp in Austin emailed about the Q3 delivery delay for the InventoryPro software, saying the November 15 deadline was missed by three weeks. ``` **After (Anonymized - Safe for Public AI):** ``` [Client Name] from [Company A] in [City] emailed about the Q3 delivery delay for the [Product Name] software, saying the [Date] deadline was missed by three weeks. ``` #### The Confidentiality Decision Tree Before using AI with any information, follow this decision tree: ``` START: Do I need to use AI for this task? ↓ YES → Does this involve client-specific information? ↓ YES → Is this information already public? ↓ NO → Can I effectively anonymize it? ↓ NO → Do we have a secure, firm-approved AI platform? ↓ NO → DO NOT USE AI Perform task manually or request firm to approve appropriate platform ↓ YES → Use only firm-approved secure platform Document platform used and date ↓ YES → Anonymize thoroughly Use public AI with anonymized version only Document what was anonymized ↓ YES (already public) → May use public AI Still exercise caution with sensitive legal strategy ↓ NO (no client-specific info) → May use public AI Still follow verification protocols ``` ### Understanding and Managing AI Hallucinations AI "hallucination" occurs when a language model generates text that is fluent and plausible but factually incorrect or entirely fabricated. In legal work, this poses extraordinary risk. #### Common Types of Legal Hallucinations **1. Fabricated Case Citations** The AI invents case names, citations, and holdings that sound real but don't exist. **Example**: * AI cites: *Smith v. Jones Enterprises, 456 F.3d 789 (7th Cir. 2018)* * Reality: This case does not exist * The citation format looks correct, making it particularly dangerous **2. Misattributed Holdings** The AI cites a real case but misstates what the case actually held. **Example**: * AI states: *Brown v. Board of Education* held that the "separate but equal" doctrine applied to public schools * Reality: *Brown v. Board of Education* **overturned** the "separate but equal" doctrine **3. Outdated or Overruled Precedent** The AI cites a case that was valid at one time but has since been overruled or superseded. **4. Fabricated Statutes or Regulations** The AI invents statutory language or regulation numbers that sound plausible. **5. Misapplied Legal Standards** The AI applies the correct legal principle but to the wrong jurisdiction or factual scenario. #### Anti-Hallucination Prompt Strategies Build these safeguards directly into your prompts: **Strategy 1: Explicit Uncertainty Instructions** ``` If you do not know the answer or do not have sufficient information to answer reliably, respond by saying "I do not have enough information to answer this question" rather than generating a speculative response. Do not fabricate case citations, statutes, or legal principles. If you are uncertain, clearly state your uncertainty. ``` **Strategy 2: Request Source Attribution and Confidence Levels** ``` For each legal proposition you state: 1. Cite the specific source (case, statute, regulation) 2. Provide your confidence level: - HIGH: Well-established principle with clear authority - MEDIUM: Generally accepted but with some variation - LOW: Uncertain or evolving area - SPECULATIVE: No clear authority; educated inference only If you cite a case, include the full Bluebook citation and a parenthetical explanation of its holding. ``` **Strategy 3: Require Explicit Acknowledgment of Limitations** ``` Before providing your analysis, explicitly identify: 1. Any gaps in the information provided 2. Any assumptions you are making 3. Any areas where the law is unsettled 4. Any jurisdiction-specific variations that may apply If you cannot find relevant authority, state: "I could not locate relevant authority on this specific issue" rather than fabricating sources. ``` **Strategy 4: Challenge the AI's Initial Response** After receiving an AI response, use a follow-up prompt to stress-test it: ``` Review your previous response carefully. Identify: 1. Any citations that you are not 100% certain exist 2. Any legal conclusions that could be challenged 3. Any alternative interpretations of the law 4. Any contrary authority that might exist Be honest about areas of uncertainty. ``` #### The Hallucination Verification Protocol For any AI-generated legal content, follow this protocol: **Step 1: Citation Check** * [ ] Every case citation independently verified in Westlaw/Lexis * [ ] Case name, reporter, and year confirmed * [ ] Parallel citations checked if available * [ ] Shepardize/KeyCite every case for validity * [ ] Verify the case actually stands for the cited proposition **Step 2: Quotation Verification** * [ ] Every quotation traced to the original source * [ ] Context of quotation reviewed * [ ] Ellipses and brackets verified as accurate * [ ] No quotations taken out of context **Step 3: Legal Principle Verification** * [ ] Legal standards cross-referenced with authoritative sources * [ ] Jurisdiction-specific rules confirmed * [ ] Recent changes in law researched * [ ] Alternative interpretations considered **Step 4: Factual Verification** * [ ] Every fact traced to source document * [ ] Page and line citations confirmed * [ ] No facts misstated or taken out of context * [ ] Dates, names, and numbers double-checked ### Mandatory AI Review Protocol (MARP) Implement this three-step protocol for reviewing any substantive legal or factual output generated with AI assistance. #### Step 1: Citation Validation **Responsibility**: Lawyer or Senior Paralegal **Action Items**: * Manually confirm the existence, currency, and relevance of *every* legal citation using a verified legal database * Use Shepard's Citations (Lexis) or KeyCite (Westlaw) to verify the citation is still good law * Confirm the case or statute actually supports the proposition for which it is cited * Check that jurisdiction and procedural posture are correctly stated **Documentation**: Create a citation verification log: | Citation | Verified In | Status | Supports Proposition? | Notes | | ------------ | ------------- | -------- | --------------------- | ----------------------------- | | \[Case name] | Westlaw | Good Law | Yes | Reviewed full opinion | | \[Statute] | Official Code | Current | Yes | Checked for recent amendments | #### Step 2: Factual Grounding **Responsibility**: Paralegal or Associate **Action Items**: * For every factual assertion, cross-reference the claim back to the source document * Ensure page and line citations are accurate * Verify quotations match the source exactly * Confirm dates, names, and numbers are correct * Check that context has not been distorted **Red Flags to Watch For**: * Factual statements without source citations * Round numbers that seem estimated rather than exact * Dates or timelines that don't align with known case chronology * Names or titles that differ from source documents * Paraphrases that change the meaning of the original #### Step 3: Output Labeling **Responsibility**: Originator (Lawyer or Paralegal) **Action Items**: Every AI-assisted draft **must** be clearly labeled on the first page or in metadata as: ``` DRAFT: AI-ASSISTED - SUBJECT TO VERIFICATION Date generated: [Date] AI model used: [Model name] Verification status: [ ] Citations verified [ ] Facts verified [ ] Attorney approved ``` This flag alerts the entire team to apply MARP before finalization. **Final Sign-Off**: Only after completing all verification steps should the label be changed to: ``` FINAL - VERIFIED AND APPROVED AI-assisted draft: Yes Verification completed by: [Name] Final approval by: [Attorney Name] Date: [Date] ``` ### Candor Toward the Tribunal Model Rule 3.3 requires candor toward the tribunal. This duty has specific implications for AI use in legal practice. #### Disclosure of AI Use in Court Filings While there is no uniform requirement to disclose the use of AI for routine tasks, several jurisdictions have adopted rules or standing orders requiring disclosure: **Courts Requiring Disclosure** (as of publication): * U.S. District Court for the Northern District of Texas (Standing Order) * U.S. District Court for the Southern District of New York (Individual judge orders) * Various state courts have issued guidance **Best Practice**: Even absent a requirement, consider including a certification statement: ``` CERTIFICATION REGARDING AI-ASSISTED RESEARCH Counsel certifies that artificial intelligence tools were used to assist in the preparation of this filing. All legal citations and authorities have been independently verified by counsel in primary legal sources. Counsel takes full responsibility for the accuracy and appropriateness of all content in this filing. Date: _______________ _______________________________ [Attorney Name] Attorney for [Party] ``` #### When AI Use Must Be Disclosed **Mandatory Disclosure Situations**: 1. **Court Orders Requiring Disclosure**: Always comply with local rules or standing orders 2. **eDiscovery Technology**: When using AI-powered tools like TAR/Predictive Coding, the methodology must be disclosed to opposing counsel and the court 3. **Expert Reports**: If AI was used to generate data, analysis, or conclusions in an expert report, this may need to be disclosed 4. **Material to the Case**: If AI use is material to an issue in the case (e.g., opposing counsel challenges your document review methodology) **Discretionary Disclosure**: * Using AI for routine research and drafting does not require disclosure * Using AI to organize exhibits or create timelines generally does not require disclosure * Use professional judgment based on local rules and case circumstances #### The Duty of Competence in AI Use Model Rule 1.1, Comment 8 states: "To maintain the requisite knowledge and skill, a lawyer should keep abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology..." **Competence Requirements**: 1. **Understanding AI Capabilities**: Know what AI can and cannot do reliably 2. **Understanding AI Limitations**: Recognize hallucination risk and other failure modes 3. **Proper Verification**: Implement robust verification protocols 4. **Staying Current**: Follow developments in AI technology and legal AI use 5. **Appropriate Supervision**: Ensure paralegals and junior attorneys using AI are properly trained **Warning Signs of Incompetent AI Use**: * Filing AI-generated content without verification * Not understanding how the AI tool works * Failing to recognize obvious AI hallucinations * Not staying current with court rules on AI disclosure * Inadequate supervision of staff using AI * Ignoring firm policies on AI use ### Practical Compliance Framework Use this framework to ensure every AI-assisted task complies with ethical requirements: #### Pre-Use Checklist Before using AI for any legal task: * [ ] **Task is appropriate for AI assistance**: Not involving final legal conclusions without verification * [ ] **Confidentiality protected**: Either using secure platform or information is properly anonymized * [ ] **Firm approval obtained**: AI platform is on approved list or task is appropriate for public AI * [ ] **Proper training completed**: User understands AI limitations and verification requirements * [ ] **Client consent obtained** (if required by firm policy): Client is aware of and approves AI use * [ ] **Supervision in place**: Appropriate oversight for paralegal or junior attorney work #### During-Use Checklist While working with AI: * [ ] **Prompts crafted carefully**: Using techniques from Chapter 2 to minimize hallucination risk * [ ] **Confidential information protected**: Not inputting privileged or confidential information into unsecured platforms * [ ] **Output treated as draft**: Understanding this is preliminary work requiring verification * [ ] **Multiple sources consulted**: Not relying solely on AI for legal conclusions * [ ] **Work documented**: Keeping records of AI use for file documentation #### Post-Use Checklist After generating AI output: * [ ] **All citations verified**: Every case, statute, and regulation checked in primary sources * [ ] **All facts verified**: Every factual assertion traced to source documents * [ ] **Legal conclusions reviewed**: Attorney has exercised independent professional judgment * [ ] **Client interests protected**: Work product serves client's best interests * [ ] **Court rules followed**: Any required disclosures made * [ ] **Proper documentation**: File contains appropriate notation of AI use and verification * [ ] **Quality meets professional standards**: Work product meets the same standard as if created entirely by humans ### Real Disciplinary Cases: Learning From Others' Mistakes Understanding real cases where lawyers faced discipline for AI misuse provides valuable lessons. #### Case Study 1: Mata v. Avianca, Inc. (S.D.N.Y. 2023) **What Happened**: Attorney Steven Schwartz used ChatGPT to conduct legal research for a brief opposing a motion to dismiss. ChatGPT generated six fake cases with fake quotes and fake internal citations. Schwartz filed the brief without verifying the citations. **The Consequences**: * Court sanctioned both the attorney and his law firm * Required to pay the opposing party's legal fees * Public reprimand and widespread media coverage * Disciplinary referral to the appropriate grievance committee **The Court's Analysis**: "Technological advances are commonplace and there is nothing inherently improper about using a reliable artificial intelligence tool for assistance. But existing rules impose a gatekeeping role on attorneys to ensure the accuracy of their filings." **Key Lessons**: 1. AI-generated citations must be independently verified 2. "I didn't know the AI would hallucinate" is not a defense 3. The duty of competence includes understanding AI limitations 4. Both the individual attorney and the firm can face sanctions 5. Public embarrassment can damage professional reputation irreparably #### Case Study 2: Park v. Kim (Supreme Court of British Columbia, 2023) **What Happened**: A lawyer submitted fake case law generated by AI in a family law case. When opposing counsel could not locate the cases, the lawyer doubled down and provided fabricated case summaries. **The Consequences**: * Finding of professional misconduct * Costs award against the lawyer * Referral to law society for discipline * Case delayed and client prejudiced **Key Lessons**: 1. Compounding the error makes it worse 2. When citations can't be verified, admit the mistake immediately 3. Don't fabricate additional information to cover up AI hallucinations 4. The duty of candor to the tribunal is non-negotiable #### Case Study 3: Multiple Cases of Inadequate Disclosure Several cases have involved attorneys who used AI-generated content without proper disclosure when court rules required it. **Key Lessons**: 1. Check local rules and standing orders before using AI 2. When disclosure is required, be transparent and complete 3. Create a firm-wide protocol for checking disclosure requirements 4. Document your compliance with disclosure requirements ### Ethical Use of AI: Best Practices Summary #### The Ten Commandments of Ethical AI Use in Legal Practice 1. **Verify Everything**: Never file or rely on AI output without independent verification 2. **Protect Confidentiality**: Use only secure, approved platforms for confidential information 3. **Understand the Technology**: Know how AI works and its limitations 4. **Maintain Professional Judgment**: AI assists; you decide 5. **Supervise Appropriately**: Ensure paralegals and junior attorneys are trained and supervised 6. **Follow Court Rules**: Comply with all disclosure requirements and standing orders 7. **Document Your Process**: Keep records of AI use and verification steps 8. **Stay Current**: Keep up with evolving AI technology and legal standards 9. **Be Transparent**: When disclosure is appropriate, be forthcoming about AI use 10. **Put Client Interests First**: Use AI to improve client service, not as a shortcut that creates risk #### When in Doubt If you're uncertain whether a particular use of AI is appropriate: 1. **Check your firm's AI policy** (if one exists) 2. **Consult with risk management or ethics counsel** 3. **Research whether your jurisdiction has issued guidance** 4. **Err on the side of caution and transparency** 5. **Document your decision-making process** ### Chapter Summary Ethical AI use in legal practice requires: * **Verification**: Every citation, fact, and legal conclusion must be independently verified * **Confidentiality**: Client information must be protected through secure platforms or proper anonymization * **Competence**: Understanding AI capabilities, limitations, and appropriate use cases * **Candor**: Transparent disclosure when required and honest recognition of AI limitations * **Professional Judgment**: Human oversight and decision-making cannot be delegated to AI * **Proper Documentation**: Clear records of AI use and verification steps The cases of sanctioned attorneys demonstrate that AI misuse has real consequences. But used responsibly, AI is a powerful tool that can enhance the quality and efficiency of legal work while maintaining the highest professional standards. Remember: AI is your assistant, not your replacement. The judgment, ethics, and professional responsibility remain entirely yours. *** *In Chapter 5, we'll explore how to build effective AI workflows that integrate these ethical guardrails into systematic, repeatable processes that improve your practice while maintaining professional standards.* # 4.1. Prompt Chaining In the context of large language models (LLMs), prompt chaining refers to a technique where multiple prompts are chained together to generate a sequence of outputs from the model. The idea behind prompt chaining is to enable the model to generate longer sequences of text or to perform more complex tasks by breaking them down into smaller sub-tasks, each of which can be handled by a separate prompt. Prompt chaining can be achieved in various ways, such as: 1. Serial prompting: This involves providing the model with a sequence of prompts, one after the other, where each prompt builds upon the previous one. For example, a first prompt might ask the model to generate a list of items, and a second prompt might ask the model to describe each item on the list. 2. Hierarchical prompting: This involves using a hierarchical structure of prompts, where a high-level prompt is broken down into lower-level prompts, each of which is further refined until the task is completed. For example, a high-level prompt might ask the model to write a story, and lower-level prompts might ask the model to generate characters, plot points, and sentences. 3. Hybrid prompting: This combines serial and hierarchical prompting techniques to create a hybrid approach. For example, a high-level prompt might ask the model to generate a report, and lower-level prompts might ask the model to provide data analysis, charts, and tables for the report. 4. Self-prompting: This involves training the model to generate its own prompts based on the input it has received so far. This allows the model to continue generating output even when the human-provided prompts run out. By chaining prompts together, LLMs can generate longer sequences of text or perform more complex tasks than they would be able to do with a single prompt. However, it's important to note that prompt chaining can also introduce new challenges, such as ensuring that the prompts are well-coordinated and that the model doesn't get stuck in an infinite loop. Researchers are actively exploring different approaches to address these challenges and improve the effectiveness of prompt chaining in LLMs. # 4.2. Output Parsers Output parsers can be used when you want to get more structured information than just text back from ChatGPT. They allow you to customize and format the text responses into different types, such as Markdown, HTML, tables, etc. Output parsers can make your life easier and cut down on the amount of copy+paste+reformatting you have to do between ChatGPT and your documents. {% content-ref url="/pages/f2p7r5jK9KGoSyiVZRPK" %} [4.2.1. Markdown](/4.-ethical-guardrails-and-professional-responsibility/4.2.-output-parsers/4.2.1.-markdown) {% endcontent-ref %} {% content-ref url="/pages/O6AikmtpeDGQVgi0w8LF" %} [4.2.2. HTML](/4.-ethical-guardrails-and-professional-responsibility/4.2.-output-parsers/4.2.2.-html) {% endcontent-ref %} {% content-ref url="/pages/SdUYAcko4IHrsfDF5kKI" %} [4.2.3. Graphviz (Dot Language)](/4.-ethical-guardrails-and-professional-responsibility/4.2.-output-parsers/4.2.3.-graphviz-dot-language) {% endcontent-ref %} # 4.2.1. Markdown {% hint style="info" %} TL;DR – copy & paste your ChatGPT responses into your documents with proper formatting by default (page headings, lists, tables, bolded text, etc.) {% endhint %} Markdown is a simple and intuitive way of writing formatted text using plain text syntax. It allows you to easily create headings, lists, tables, and links. Markdown is widely used in certain industries, but has yet to adopted by the legal industry. We believe this will quickly change as ChatGPT and other generative AI tools become widely uses by legal professionals. ### Two reasons why we like using Markdown with ChatGPT: #### 1. Increasing readability of ChatGPT responses If we provide a simple prompt that simply asks ChatGPT to summarize the Securities Act of 1933, we'll get a response that looks similar to: > The Securities Act of 1933, also known as the "truth in securities" law, was the first significant federal legislation to regulate the offer and sale of securities. It was enacted in the aftermath of the stock market crash of 1929 to restore investor confidence in the capital markets by providing more structure and government oversight. > > **Section 2: Definitions** > > 1. *Highlight*: This section defines the fundamental terminologies used throughout the Act, including "security," "issuer," "underwriter," and "prospectus." This clarification aids in the precise interpretation and application of the Act. > 2. *Highlight*: The broad definition of "security" given in this section has been the basis for the Act's extensive applicability. It covers a range of financial instruments, from traditional stocks and bonds to more complex and exotic derivatives. > > ... ChatGPT has done a good job at formatting the response to make it legible. But, we can do better. Let's try adding the following sentence to the original prompt asking ChatGPT to summarize the Securities Act of 1933: ``` Format the summary using markdown elements such as headings, lists, blockquotes, and tables where appropriate. ``` This simple addition makes a noticeable different to the response returned by ChatGPT...

ChatGPT returns a nicely formatted response using Markdown

#### 2. Streamlining copy+paste of ChatGPT responses into your documents with formatting What we saw in our first reason above is "rendered" Markdown, which is what we want when we only wish to view a nicely formatted response within ChatGPT. However, if we want to copy+paste the response into our word document it is easier to work with the "raw" Markdown. We can have ChatGPT return the "raw" Markdown by simply including `"Return the raw markdown"` or something to that tune in our prompt. Using the same example as before, ChatGPT's response would look something like: 
# The Securities Act of 1933: An Overview

The Securities Act of 1933, also known as the "truth in securities" law, was the first major federal legislation to regulate the offer and sale of securities. It was enacted following the stock market crash of 1929, with the aim of restoring investor confidence in the capital markets by increasing government oversight and transparency.

## Section 2: Definitions

- **Highlight 1**: This section provides definitions for key terms used throughout the Act, such as "security," "issuer," "underwriter," and "prospectus." These definitions assist in the interpretation and application of the Act.
...
If you want to learn more about the Markdown syntax elements seen in the response above ("#", "\*\*text\*\*") take a look at the [Markdown Cheat Sheet](https://www.markdownguide.org/cheat-sheet/). Copying and pasting the raw Markdown into popular word editing software will automatically format the text according to the Markdown rules. Gone are the days of you having to perform tedious reformatting :relaxed:. [^1]: Marks the start of a heading. [^2]: The "##" syntax element is used to create subheadings. [^3]: \*\*\\*\* is used to bold text. # 4.2.2. HTML {% hint style="info" %} TL;DR – Useful when planning on taking ChatGPT's responses and including them on your firm's website or blog post(s). {% endhint %} HTML is the language used to create and structure webpages. For lawyers and paralegals, having ChatGPT return its responses in HTML can simplify the process of updating your firm's website or blog. Also, some case management platforms, and other legal software tools include a [WYSIWYG](https://en.wikipedia.org/wiki/WYSIWYG) editor allowing you to create legal documents that are well-formatted, easily navigable, and visually appealing. Hence, having ChatGPT output its responses in HTML could save you a fair amount of copy+paste+formatting work. Here's a snippet of a HTML response ChatGPT provided us with: ```html

The Securities Act of 1933

The Securities Act of 1933, known as the "truth in securities" law, aimed to restore investor confidence in the capital markets by providing more structure and government oversight.

Section 2: Definitions

... ``` Having ChatGPT return its results in HTML format is as simple as adding \ `"Provide your response in HTML."` within your prompt. # 4.2.3. Graphviz (Dot Language) Graphviz is a freely available tool for creating visual diagrams, ranging from simple entity-relationship diagrams to more complex flowcharts. Graphviz includes a special text-based syntax that can be used to programmatically draw diagrams called Dot Language. This means we can have ChatGPT create diagrams for us by asking it provide its response in Dot Language. We can then take the generated Dot Language text and use Graphviz or other common software tools to generate an image of the diagram. We saw ChatGPT output responses in Dot Language back in [section 2.1.3](/2.-fundamentals-of-legal-prompt-engineering/2.1.-prompt-templates/2.1.3.-extraction) to generate an organizational chart. Now let's put on our estate planning hats and work through a prompt that will have ChatGPT create a visual representation for a hypothetical client in need of business & estate planning guidance. ### Backstory (fictional) Meet the Walker family from Dallas, Texas. The Walker's have a growing real estate portfolio consisting of residential rental units. Johnny Walker, the wealth creator of the family, and his wife Jessica, are ready to bring their two children into the business to help manage the properties. We won't get into the semantics, but their imaginary attorney, Mark Case, rendered hypothetical counsel on how they should organize their holdings using a Series LLC. Mark Case is prompt engineering wizard and had ChatGPT generate a diagram illustrating what this might look like using Dot language. ### Our Prompt: ``` Using Dot language, create a diagram for a Series LLC. The master LLC will be named "Walker Real Estate Holdings LLC" and 7 series (children) named to match this pattern: "Walker Real Estate Holdings LLC - Series {number}" where {number} is replaced with the number of the series. ``` As expected, ChatGPT returns a diagram output in Dot language that we can use to generate an image that can be shared with the Walker family.

This image was generated via ChatGPT's response

You can take a look at our ChatGPT interaction ([link](https://chat.openai.com/share/69444df7-2667-4e70-a451-2d2ba8230d5d)) to view the Dot language response. Don't be afraid to tinker with this example by modifying our prompt, and submitting iterative prompts with additional information for further learning. ### Thought Experiment: I got sidetracked Estate planning has always been an area of law of interest to me. I wanted to find out how strong ChatGPT's legal intuition is, if one even exists. I took a step back from the prompt we used above and decided to provide a minimal amount of information regarding the Walker's family situation without prescribing the legal strategy that should be taken (organize a Series LLC). It's response was better than I expected. Take a look at my conversation with ChatGPT [here](https://chat.openai.com/share/df20a9d1-023c-4f54-b08f-ec1c0787a2fa). I had to guide the AI model to go down the path of architecting a Series LLC. I also had to revise my prompt once to instruct it to derive the name of each Series LLC from the Master's name (I momentarily forgot about the importance of specificity :smile:). # 4.3. ChatGPT Plugins ChatGPT plugins are extensions that enhance the functionality and performance of ChatGPT. They connect ChatGPT to external applications, enabling it to fetch real-time information, access knowledge bases, and assist with tasks like scheduling and document drafting. Plugins make ChatGPT a valuable and efficient tool for legal professionals. {% hint style="info" %} **You must be a ChatGPT Plus user to access Plugins. To learn how to access plugins within your OpenAI ChatGPT account reference the link** [**here**](https://help.openai.com/en/articles/7183286-how-do-i-access-plugins)**.** {% endhint %} In this section we're going to go over a few of the most popular plugins currently available. We're especially interested in leveraging plugins that provide us with the following capabilities: * Accessing the internet to retrieve up to date information. * Engage with PDFs or Word documents for tasks such as summarizing content, extracting information, and posing questions. * Accessing external applications (ie. Email, Calendar, Case Management Platforms, etc.) to automate simple tasks ## Getting Started Let's first make sure we have Plugins enabled. Then we'll browse the available plugins we can leverage within ChatGPT. Take a look at the demo below. {% @arcade/embed flowId="d4MhGam0eaTasf5GNzUv" url="" %} ## Let's start using Plugins! ### Example: ChatWithPDF Being able to have ChatGPT work with PDF documents is invaluable for many. Being able to summarize large PDFs, produce answers using the contents of the PDF for any questions you pose, and more solves endless use cases for legal professionals. Before we dive into our example, we can emphasize enough the importance of understanding how services like ChatGPT and plugins process the contents of your document. If you're looking for an AI service that securely and privately handles PDF documents take a look at [CaseMark AI](https://casemark.ai). {% hint style="danger" %} WARNING: Do **NOT** upload PDF files that have sensitive or proprietary information. Doing so could result in data leakage and privacy violations. {% endhint %} Diving in: I recently read a research paper which explored the capabilities of LLMs in applying tax law. We'll have the ChatWithPDF plugin download this PDF and reference it as we ask ChatGPT questions on the research that was conducted. Take a look at our interactive demo below. {% @arcade/embed flowId="OJkH4SAsx26Ip2QBDwW1" url="" %} As you can see, it's fairly straightforward to start interacting with your PDFs. We can also provide multiple PDFs in the same chat session as can be seen in the screenshot below. When working with multiple PDF documents, ChatGPT will intelligently reference the related document(s) when generating a response to the prompt you provided.

ChatWithPDF Plugin: Working with multiple documents

### Example: Connecting ChatGPT to the internet with "Web Requests" Odds are that if you've ever used ChatGPT in any capacity you've likely come across one of its infamous responses – "As an AI language model, I have a knowledge cutoff date because my training data only goes up until September 2021." The GPT in "ChatGPT" stands for **Generative Pre-trained Transformer,** meaning that it was trained on a massive amount of existing text data scraped from the internet. Any data made available on the internet after September 2021 was not used to train ChatGPT. Therefore, it has no knowledge of recent events, and it has no connection to the outside world. ChatGPT can NOT tell you what's the current temperature in Miami, FL. However, ChatGPT with a plugin that has access to the internet can solve this problem. In the interactive demo below we walk you through the process of installing and using a plugin that can search the internet for ChatGPT – "Web Requests." {% @arcade/embed flowId="l02GMa4SxAczRKFK1Ib8" url="" %} You've now seen how you can leverage plugins to interact with PDF documents, and search the web to gain real-time access to information. If you're asking yourself whether you can use multiple plugins at the same time you're on the right track 🙂. Our next example will demonstrate how you can have multiple plugins enabled at the same time, extending the capabilities of ChatGPT and helping you tackle more complex tasks. ### Example: Using multiple Plugins simultaneously Let's supercharge our ChatGPT workflow by using multiple plugins in the same chat session. Take a look at our interactive demo below that demonstrates how we expanded on one of our earlier examples which leveraged the ChatWithPDF plugin to extract information from a research paper. This time around we enabled the "Web Requests" plugin to give ChatGPT access to the internet. We proceeded by having ChatGPT return any amendments made to the Internal Revenue Code in 2023. {% @arcade/embed flowId="SZVtm2bcs17Uafz4bp9c" url="" %} ### What's next? We encourage you to explore the hundreds of available ChatGPT plugins to see which ones can help make you more productive when using ChatGPT. In early July 2023, OpenAI announced Code Interpreter which many claim is the next evolution of plugins. We'll take a look at Code Interpreter in the next section. # 4.4. ChatGPT Code Interpreter Some folks on the internet (ourselves included) are calling OpenAI's Code Interpreter a "game changer" and "the next evolution of ChatGPT capabilities." ChatGPT, besides being able to write code like an experienced computer programmer, can now execute that code for you thanks to Code Interpreter. This paves the wave for you to complete more advanced use-cases for using ChatGPT. For example, an estate planning attorney can upload a template of a Trust contract and have ChatGPT fill it out using the clients information. Or, a tax attorney can upload a spreadsheet of a client's stock trades and have Code Interpreter calculate capital gains and other relevant figures and export the results. A paralegal can have a client intake form generated as a PDF in seconds. There are countless scenarios where legal professionals might find the capabilities of Code Interpreter useful in accomplishing different tasks. For the more technical reader, here is how OpenAI described it in their announcement in July 2023... > *An experimental ChatGPT model that can use* [*Python*](#user-content-fn-1)[^1]*, handle uploads and downloads.*
> > *We provide our models with a working* [*Python interpreter in a sandboxed, firewalled execution environment*](#user-content-fn-2)[^2]*, along with some ephemeral disk space. Code run by our interpreter plugin is evaluated in a persistent session that is alive for the duration of a chat conversation (with an upper-bound timeout) and subsequent calls can build on top of each other.*
> > *We support uploading files to the current conversation workspace and downloading the results of your work.* Let's go over one of the example use-cases we mentioned above using Code Interpreter. We previously used ChatGPT plugins to be able to upload a PDF and have ChatGPT answer questions and extract information from the document. In this example we'll have Code Interpreter generate a downloadable PDF of a client intake form commonly sent to new clients of law firms. {% @arcade/embed flowId="nRYaPFHVFI4gE58oc0gN" url="" %} [^1]: Programming language heavily used in data science and other technology sectors. [^2]: ChatGPT can now write software, and then execute it for you. # 5. Building Effective AI Workflows Once you understand the fundamentals and ethics, this chapter shows you how to integrate AI systematically into your practice. We'll explore how to map AI use to each phase of litigation, create multi-step workflows, implement quality assurance systems, and manage costs effectively. ### Integrating AI Into Your Litigation Timeline Effective AI integration is not a single action but a set of practices applied at strategic points during case preparation. This section maps AI applications to the natural litigation workflow. #### Early Case Assessment and Strategy **Goal**: Quickly understand the scope, risk, and likely battlegrounds of the case. **AI Applications**: **Sentiment and Topic Analysis**: AI scans initial document collections (e.g., client's hard drive, key custodian emails) to identify prevailing themes, key players, and emotional tone related to the dispute. **Paralegal Example**: Use AI to run a quick topic model on 10,000 corporate emails to identify the top five subjects discussed in the three months leading up to the alleged breach. This helps prioritize which custodians to interview first and focus the initial discovery requests. **Prompt Template:** ``` **INSTRUCTIONS** Act as an eDiscovery analyst conducting early case assessment. Analyze the document collection to identify key themes, players, and patterns that will inform litigation strategy. **CONTEXT** Patent infringement case. We represent the plaintiff alleging defendant copied our proprietary software architecture. Initial discovery has produced 10,000 emails from defendant's engineering team during the relevant time period (January 2023 - June 2024). **INPUT** Analyze the attached email collection and identify: 1. Top 5 most frequently discussed topics 2. Key individuals (top 10 most active communicators) 3. Timeline of significant events based on email activity spikes 4. Emotional tone patterns (cooperative vs. adversarial) 5. Any references to our client's technology or products **OUTPUT** Format as Early Case Assessment Report: EXECUTIVE SUMMARY [2-3 paragraphs on key findings] TOPIC ANALYSIS Topic 1: [Name] - Frequency: [X emails, Y% of collection] - Key participants: [Names] - Significance: [Why this matters for the case] [Continue for all 5 topics...] KEY PLAYERS ANALYSIS [Name], [Title] - Email volume: [X sent, Y received] - Key topics: [List] - Relationship to claims: [Description] TIMELINE OF SIGNIFICANT EVENTS [Date range]: [Event description based on email patterns] STRATEGIC RECOMMENDATIONS - Priority custodians for deposition - Key search terms for next review phase - Potential weaknesses in our case - Opportunities for favorable discovery ``` **Lawyer Example**: Input the initial complaint and answer into an AI legal research tool and ask it to cross-reference the asserted causes of action against relevant jury instructions in that jurisdiction, anticipating necessary proof elements early on. **Prompt Template:** ``` **INSTRUCTIONS** You are a litigation consultant conducting preliminary legal analysis. Identify the elements that must be proven for each claim and cross-reference with applicable jury instructions. **CONTEXT** Commercial litigation in state court (California). We represent plaintiff in a breach of contract and fraud case. **INPUT** Review the attached Complaint and Answer. For each cause of action alleged: 1. Identify the legal elements that must be proven 2. Cross-reference with California Civil Jury Instructions (CACI) 3. Identify which elements are disputed vs. undisputed 4. Flag any affirmative defenses that will require additional proof **OUTPUT** Format as Legal Element Analysis: CAUSE OF ACTION #1: Breach of Contract Required Elements: 1. [Element] - CACI [number] Status: [Disputed/Undisputed] Defendant's position: [From Answer] 2. [Element] - CACI [number] Status: [Disputed/Undisputed] Defendant's position: [From Answer] Evidence Needed to Prove: - [Element 1]: [Types of evidence] - [Element 2]: [Types of evidence] Affirmative Defenses to Address: - [Defense]: Elements and burden Discovery Priorities: - [Specific discovery needed] [Repeat for each cause of action...] TRIAL STRATEGY IMPLICATIONS [Summary of proof challenges and opportunities] ``` #### Discovery Phase: Document Review and Production **Goal**: Increase the speed, consistency, and accuracy of massive document review tasks. **AI Applications**: **Technology Assisted Review (TAR) / Predictive Coding**: Use machine learning to prioritize documents most likely to be relevant, privileged, or responsive to a request. **Paralegal Example**: A paralegal is tasked with training the AI. They review 500 documents and label them as "Responsive" or "Not Responsive." The AI then uses this training set to score the remaining 500,000 documents, allowing the paralegal to focus review efforts on the top 10% highest-scoring documents, saving massive amounts of time and budget. **Initial Training Prompt:** ``` **INSTRUCTIONS** Act as a document review consultant analyzing the training set results to optimize our TAR protocol. **CONTEXT** Products liability litigation. Document universe: 500,000 emails and documents. We need to identify all communications regarding product safety testing and consumer complaints. Training set: 500 documents (250 coded Responsive, 250 coded Not Responsive) **INPUT** Analyze the training set coding decisions and identify: 1. Common characteristics of responsive documents 2. Common characteristics of non-responsive documents 3. Keywords/phrases strongly associated with responsiveness 4. Custodians with high responsive document rates 5. Date ranges with higher relevance rates 6. Document types (email vs. memo vs. report) correlation with relevance **OUTPUT** Format as TAR Optimization Report: RESPONSIVE DOCUMENT PATTERNS Keywords (appear in >30% of responsive docs): - [Keyword]: [X% of responsive docs, Y% of non-responsive docs] - Analysis: [Why this matters] Custodian Analysis: - High-value: [Names with >50% responsive rate] - Low-value: [Names with <10% responsive rate] Document Type Analysis: - Emails: [X% responsive rate] - Reports: [Y% responsive rate] - Memos: [Z% responsive rate] RECOMMENDED SEARCH REFINEMENTS Boolean Search String: [Suggested search based on patterns] NEXT TRAINING ROUND RECOMMENDATIONS - Sample size: [Number of documents] - Focus areas: [Specific custodians/date ranges] - Quality control measures: [Suggestions] ``` **Privilege Log Generation**: Use AI to identify documents containing attorney email domains and legal terminology, creating a preliminary privilege log for attorney review. **Prompt Template:** ``` **INSTRUCTIONS** Act as a privilege review coordinator. Your role is to FLAG potentially privileged documents for attorney review, not make final privilege determinations. Be conservative. Flag any document that might have a colorable privilege claim. **CONTEXT** Employment litigation. We represent the employer. Producing 50,000 documents in response to plaintiff's requests. Known attorney identifiers: - Outside counsel: @employmentlawfirm.com - In-house counsel: [Names], legal@company.com - Keywords: "privileged," "attorney-client," "work product" **INPUT** Review documents and identify those requiring privilege review. For each flagged document, note: 1. Document ID 2. Privilege indicators present (attorney email, legal keywords, etc.) 3. Type of privilege potentially applicable 4. Level of confidence (High/Medium/Low) **OUTPUT** Format as Privilege Review Queue: HIGH PRIORITY FOR REVIEW (strong privilege indicators) Doc ID: [Number] From: [Attorney email domain identified] To: [Recipients] Date: [Date] Subject: [Subject line] Indicators: [What triggered the flag] Privilege Type: [Attorney-client / Work product / Both] MEDIUM PRIORITY FOR REVIEW (possible privilege indicators) [Same format] LOW PRIORITY FOR REVIEW (marginal indicators) [Same format] SUMMARY STATISTICS Total documents reviewed: [X] High priority: [Y] (Z%) Medium priority: [Y] (Z%) Low priority: [Y] (Z%) Clear non-privileged: [Y] (Z%) QUALITY CONTROL NOTES [Any patterns or issues noticed during review] ``` #### Pre-Trial Phase: Motion Practice and Witness Preparation **Goal**: Draft high-quality filings efficiently and prepare witnesses thoroughly. **AI Applications**: **Fact Synthesis and Cross-Referencing**: AI tools can connect scattered data points across transcripts and exhibits. **Paralegal Example**: A paralegal needs to prepare a summary of inconsistencies in a key witness's testimony. They use AI to query all 15 exhibits and 3 deposition transcripts, asking for all dates mentioned by the witness regarding the product launch compared to dates in the exhibits. **Prompt Template:** ``` **INSTRUCTIONS** Act as a trial preparation paralegal creating impeachment materials. Identify inconsistencies between witness testimony and documentary evidence with precise citations. **CONTEXT** Products liability case. Key defense witness is the product manager who claims the safety testing was completed before product launch. We have: - Witness deposition (200 pages) - 15 exhibits (safety test reports, launch timeline documents, emails) **INPUT** Cross-reference the deposition testimony with the exhibits to identify: 1. Dates the witness stated for key events 2. Dates shown in documentary evidence for the same events 3. Any contradictions or inconsistencies 4. Statements the witness made that are contradicted by documents Focus on: - Product launch date - Safety testing completion dates - Approval timeline - When witness became aware of safety issues **OUTPUT** Format as Impeachment Chart: TOPIC: Product Launch Date Witness Testimony: "We launched in June 2023" [Tr. 45:12-15] Documentary Evidence: Exhibit 7: Email dated May 15, 2023 with subject "Launch Next Week" Exhibit 12: Press release dated May 22, 2023 announcing product availability Inconsistency: Witness testified launch was June 2023, but documents show May 2023 launch. Discrepancy: Approximately 2-4 weeks Significance: If testing was completed "just before launch" as witness testified, the actual timeline was more compressed than he acknowledged. [Continue for each identified inconsistency...] SUMMARY OF INCONSISTENCIES Total identified: [X] Major (materially affect credibility): [Y] Minor (timing/detail issues): [Z] IMPEACHMENT STRATEGY RECOMMENDATIONS Priority order for cross-examination: 1. [Most significant inconsistency] 2. [Second most significant] [etc.] ``` **Boilerplate Drafting**: Use AI to generate standard sections of motions (e.g., standard of review, jurisdictional statement), allowing the lawyer to focus on substantive legal arguments. **Prompt Template:** ``` **INSTRUCTIONS** Draft standard procedural sections for a motion brief. These sections should be professionally written but do not require novel legal analysis. **CONTEXT** Federal district court (Central District of California) Motion type: Motion for Summary Judgment Case type: Employment discrimination (Title VII) **INPUT** Draft the following standard sections: 1. Caption (use placeholder [CASE NAME]) 2. Table of Contents 3. Table of Authorities (leave entries blank for insertion) 4. Introduction (1 paragraph overview - use placeholder [BRIEF DESCRIPTION]) 5. Statement of Jurisdiction 6. Standard of Review for Summary Judgment in federal court 7. Statement of Undisputed Facts (format only, with instructions for completion) 8. Conclusion and Prayer for Relief **OUTPUT** Format as formal motion brief with proper formatting: [Include proper heading format, section numbering, etc.] Focus on: - Professional legal writing - Proper citation format (Bluebook) - Appropriate headings and subheadings - Correct procedural standards for this court and motion type Note: Substantive legal arguments will be added separately by attorney. ``` #### Trial Phase: Real-Time Support **Goal**: Provide quick access to information and documents during trial. **AI Applications**: **Document Retrieval**: Quickly locate exhibits, deposition testimony, or legal authority during trial. **Prompt Template (Pre-Trial Setup):** ``` **INSTRUCTIONS** Create a comprehensive trial reference system that enables rapid retrieval of information during trial. **CONTEXT** Construction defect trial. 3-week trial starting next month. We have: - 150 exhibits - 12 deposition transcripts - 50+ legal authorities we may cite **INPUT** Create searchable reference guides for: 1. Exhibit quick-reference (by topic, witness, chronology) 2. Deposition testimony index (by topic and witness) 3. Legal authority quick-cite (by legal issue) **OUTPUT** Format as three separate quick-reference guides: EXHIBIT QUICK-REFERENCE GUIDE By Topic: CONTRACT FORMATION - Exhibit 1: Original contract [Key provisions: pp. 3-5] - Exhibit 2: Change Order #1 [Added bathroom] - Exhibit 5: Email confirming terms [Date: 3/15/23] DEFECT EVIDENCE - FOUNDATION - Exhibit 20-35: Photos of foundation cracks - Most dramatic: Ex. 23, Ex. 28 - Show progression: Ex. 20 (first), Ex. 35 (worst) [Continue by topic...] DEPOSITION TESTIMONY INDEX WITNESS: John Smith (Contractor) Topic: Knowledge of Building Codes - 45:10-47:3: Admits he knew code required X - 112:5-113:22: Cannot explain why he used Y instead - Cross-ref: Exhibit 62 (actual code provision) Topic: Timeline Issues - 78:12-80:5: Claims work took longer due to weather - Impeachment: Exhibit 40 (weather records show minimal rain) [Continue for each witness and topic...] LEGAL AUTHORITY QUICK-CITE ISSUE: Breach of Contract - Substantial Performance Primary: [Case name], [Citation] - Holding: [One sentence] - Key quote: [With page number] - Use for: [When this applies] Alternative: [Second case] [Same format] [Continue for each legal issue...] ``` ### Multi-Step Workflows for Complex Tasks Complex legal projects benefit from using AI at multiple stages, with each output feeding into the next phase of work. #### Comprehensive Litigation Matter Analysis Workflow This workflow demonstrates how to use AI strategically at different stages of case development: **Stage 1: Initial Document Processing** ``` **INSTRUCTIONS** Act as a case assessment analyst conducting initial document review. **CONTEXT** New case intake. Potential client brings 200 documents (emails, contracts, financial records) related to a business dispute. **INPUT** Process all documents and create: 1. Chronological timeline of key events 2. Identification of key players and their roles 3. Document categories (contracts, correspondence, financial) 4. Preliminary assessment of strengths and weaknesses **OUTPUT** [Detailed analysis as outlined above] Save this output as "Stage 1 Analysis" for use in Stage 2. ``` **Stage 2: Legal Theory Development** (Uses Stage 1 output) ``` **INSTRUCTIONS** You are a litigation strategist. Using the case facts and documents identified in Stage 1, develop potential legal theories. **CONTEXT** Business dispute. Stage 1 Analysis identified: - Breach of partnership agreement (dated 1/15/23) - Alleged misappropriation of business opportunities - Financial irregularities in partnership accounting - Key players: Partner A, Partner B, Company CFO [Paste Stage 1 Analysis results] **INPUT** Develop comprehensive litigation strategy including: 1. Potential causes of action 2. Required elements for each claim 3. Evidence available to support each element 4. Anticipated defenses 5. Discovery priorities 6. Settlement leverage points **OUTPUT** [Detailed legal strategy memo] Save this output as "Stage 2 Strategy" for use in Stage 3. ``` **Stage 3: Discovery Planning** (Uses Stage 1 & 2 outputs) ``` **INSTRUCTIONS** You are a discovery specialist. Using the legal theories from Stage 2 and the document analysis from Stage 1, create a comprehensive discovery plan. **CONTEXT** [Paste relevant portions of Stage 1 and Stage 2 outputs] **INPUT** Draft discovery plan including: 1. Priority interrogatories (15 total) 2. Key requests for production (20 total) 3. Deposition priorities (identify top 5 witnesses) 4. Third-party subpoenas needed 5. Expert witness needs 6. Timeline for discovery phases **OUTPUT** [Detailed discovery plan] ``` **Stage 4: Document Drafting** (Uses all previous outputs) ``` **INSTRUCTIONS** You are a litigation attorney drafting the initial complaint. **CONTEXT** [Paste relevant portions of Stage 1, 2, and 3 outputs] Jurisdiction: [State] Superior Court Venue: [County] Plaintiff: [Name] Defendant: [Name] **INPUT** Draft a comprehensive complaint including: 1. Caption 2. Parties 3. Jurisdiction and Venue 4. General Allegations 5. Causes of Action (based on Stage 2 analysis) 6. Prayer for Relief 7. Jury Demand **OUTPUT** [Complete complaint draft] Note: This is a first draft requiring attorney review, revision, and verification of all factual allegations against source documents. ``` #### Cost Analysis: Multi-Stage Workflow Efficiency **Traditional Approach** (all manual): * Stage 1: 10 hours paralegal time = $1,500 * Stage 2: 8 hours attorney time = $3,200 * Stage 3: 6 hours paralegal time = $900 * Stage 4: 12 hours attorney time = $4,800 * **Total: 36 hours, $10,400** **AI-Assisted Workflow**: * Stage 1: 2 hours paralegal + AI = $300 * Stage 2: 3 hours attorney + AI = $1,200 * Stage 3: 2 hours paralegal + AI = $300 * Stage 4: 5 hours attorney + AI = $2,000 * **Total: 12 hours, $3,800** **Savings: 24 hours (67%), $6,600 (63%)** ### Prompt Chaining for Complex Analysis Prompt chaining involves breaking complex legal tasks into sequential prompts that build upon each other, creating a logical flow of analysis. #### Contract Negotiation Strategy Chain **Prompt 1: Document Analysis** ``` **INSTRUCTIONS** Act as a contracts attorney conducting detailed contract review. **CONTEXT** Commercial lease negotiation. We represent the tenant. Landlord has provided their standard form lease. **INPUT** Analyze the attached lease agreement for: 1. Key business terms (rent, term, renewal options) 2. Risk allocation (liability, insurance, indemnification) 3. Unusual or concerning provisions 4. Missing standard protections for tenants 5. Areas favorable to landlord that should be negotiated **OUTPUT** Format as detailed contract analysis with specific clause references. [After receiving this output, use it in Prompt 2] ``` **Prompt 2: Market Research** (Uses Prompt 1 output) ``` **INSTRUCTIONS** You are a commercial real estate specialist. Using the contract analysis from Prompt 1, research standard market terms and negotiation strategies. **CONTEXT** [Paste key findings from Prompt 1] Market: Office space in [City] Property type: Class A office building Lease size: 10,000 sq ft Term: 5 years **INPUT** For each concerning provision identified in Prompt 1, research: 1. Standard market terms for this type of lease 2. Typical landlord vs. tenant positions 3. Industry best practices 4. Negotiation leverage points 5. Acceptable compromise positions **OUTPUT** Format as market analysis by provision with specific recommendations. [After receiving this output, use it in Prompt 3] ``` **Prompt 3: Strategy Development** (Uses Prompt 1 & 2 outputs) ``` **INSTRUCTIONS** You are a negotiation strategist preparing for lease negotiations. **CONTEXT** Contract Analysis: [Paste Prompt 1 key points] Market Research: [Paste Prompt 2 key points] Client priorities: - Must have: [List] - Important but negotiable: [List] - Nice to have: [List] **INPUT** Develop comprehensive negotiation strategy including: 1. Priority issues (must-change provisions) 2. Secondary issues (should-change provisions) 3. Tertiary issues (nice-to-change provisions) 4. Deal-breaker issues 5. Acceptable fallback positions 6. Creative solutions for impasse issues 7. Negotiation sequence (what to address first) 8. Walk-away criteria **OUTPUT** Format as negotiation strategy memo with specific tactics for each issue. ``` #### Legal Research Chain with Verification **Prompt 1: Initial Research** ``` **INSTRUCTIONS** You are a legal researcher analyzing [Legal Issue]. Provide comprehensive analysis with citations. Flag any areas of uncertainty. **CONTEXT** [Case context] **INPUT** Research: [Specific legal question] **OUTPUT** [Research memo with citations] ``` **Prompt 2: Self-Critique** (Uses Prompt 1 output) ``` **INSTRUCTIONS** Review your previous research response critically. You are now acting as a senior partner reviewing an associate's work. **INPUT** Identify in your previous response: 1. Any citations you are not 100% certain exist 2. Any legal conclusions that could be challenged 3. Any alternative interpretations not addressed 4. Any potentially contrary authority not discussed 5. Any gaps in the analysis **OUTPUT** Format as critical review with specific concerns identified. ``` **Prompt 3: Refined Analysis** (Uses Prompts 1 & 2) ``` **INSTRUCTIONS** Based on your initial research and self-critique, provide a refined analysis that addresses the concerns identified. **CONTEXT** Initial Research: [Paste Prompt 1] Self-Critique: [Paste Prompt 2] **INPUT** Provide refined analysis that: 1. Addresses gaps identified in self-critique 2. Strengthens weak arguments 3. Acknowledges contrary authority 4. Provides more nuanced conclusions 5. Adds appropriate qualifications **OUTPUT** Format as final research memorandum with enhanced analysis. Note: All citations still require manual verification in Westlaw/Lexis. ``` ### Quality Assurance Systems Implement these quality control measures to catch errors before they become problems: #### The Triple-Check System **Check 1: AI Self-Review** ``` **INSTRUCTIONS** Review your previous response for accuracy and completeness. **INPUT** [Paste AI's previous response] Evaluate: 1. Legal accuracy - are all legal statements correct? 2. Citation accuracy - do all cited cases exist and support the propositions? 3. Logical consistency - does the analysis flow logically? 4. Completeness - are there missing considerations? 5. Appropriate qualifications - are certainty levels appropriate? **OUTPUT** Format as self-review with specific concerns flagged. ``` **Check 2: Cross-Model Verification** Use a different AI model to verify critical conclusions: ``` **INSTRUCTIONS** You are a quality control reviewer examining legal analysis for accuracy. **INPUT** Review the following legal analysis: [Paste original AI response] Focus on: 1. Whether cited cases actually support the conclusions 2. Whether the legal reasoning is sound 3. Whether any contrary authority is missing 4. Whether conclusions are appropriately qualified **OUTPUT** Identify any potential issues with specific references to the analysis. ``` **Check 3: Human Review Checklist** Before finalizing any AI-assisted work, complete this checklist: * [ ] All citations verified in primary sources * [ ] All factual assertions traced to source documents * [ ] Legal conclusions reviewed by attorney * [ ] Analysis addresses client's specific situation * [ ] Appropriate disclaimers and qualifications included * [ ] Work product meets professional standards * [ ] Ethical requirements satisfied * [ ] Confidentiality protected * [ ] Court rules followed (if applicable) * [ ] Documentation complete #### Document Review Quality Control Protocol For document review projects using AI: **Phase 1: Initial Validation** * Review random sample of 100 AI-coded documents * Calculate accuracy rate * If < 75% accurate, retrain AI * If > 75% accurate, proceed to Phase 2 **Phase 2: Ongoing Monitoring** * Review 50 randomly selected documents per 5,000 reviewed * Track accuracy metrics * Adjust if accuracy drops below threshold **Phase 3: Final Validation** * Review all documents coded as "highly relevant" (top 10%) * Review random sample of "not relevant" documents * Document final accuracy metrics **Quality Metrics Template:** ``` DOCUMENT REVIEW QUALITY METRICS Project: [Name] Review Period: [Dates] Total Documents: [X] AI-ASSISTED REVIEW ACCURACY Sample size: [X documents] Correct classifications: [Y] Accuracy rate: [Z%] BREAKDOWN BY CATEGORY Responsive documents: - AI correct: [X/Y] = [Z%] Privileged documents: - AI correct: [X/Y] = [Z%] Not relevant documents: - AI correct: [X/Y] = [Z%] ERROR ANALYSIS Type of errors: - False positives: [X] ([Y%]) - False negatives: [X] ([Y%]) Pattern of errors: [Description of common error types] CORRECTIVE ACTIONS TAKEN [List of adjustments made] FINAL VALIDATION [Results of final quality check] ``` ### Cost Management and ROI Optimization Understanding and managing AI costs ensures sustainable integration into your practice. #### Cost Tracking Framework **Track These Metrics**: 1. **Time Saved**: Hours saved per task type 2. **Quality Improvement**: Error rates before/after AI 3. **Direct Costs**: AI platform subscription or API costs 4. **Indirect Costs**: Training time, verification time 5. **Client Satisfaction**: Feedback on turnaround and quality **Cost Tracking Template:** ``` MONTHLY AI USE REPORT Month: [Month/Year] TIME METRICS Task Type | Manual Time | AI-Assisted Time | Hours Saved | Value (@billing rate) ----------|-------------|------------------|-------------|--------------------- Discovery review | 120 hrs | 40 hrs | 80 hrs | $12,000 Legal research | 30 hrs | 12 hrs | 18 hrs | $5,400 Document drafting | 25 hrs | 10 hrs | 15 hrs | $3,750 Total | 175 hrs | 62 hrs | 113 hrs | $21,150 COST METRICS AI platform subscription: $500 Additional verification time: 15 hrs × $150/hr = $2,250 Training time: 5 hrs × $150/hr = $750 Total AI costs: $3,500 NET BENEFIT Time savings value: $21,150 Less AI costs: -$3,500 Net monthly benefit: $17,650 ROI: 504% QUALITY METRICS Tasks with errors (pre-AI): 8% error rate Tasks with errors (post-AI): 2% error rate Improvement: 75% reduction in errors CLIENT FEEDBACK Turnaround time satisfaction: Improved 40% Work quality satisfaction: Maintained Cost satisfaction: Improved 35% ``` #### ROI Calculation Formula ``` ROI = (Time Saved × Billable Rate - AI Costs - Training/Verification Costs) / Total Investment × 100 Example for Solo Practitioner: - Time saved: 40 hours/month - Billing rate: $300/hour - Time savings value: $12,000/month - AI subscription: $200/month - Verification time: 8 hours × $150/hour = $1,200/month - Monthly investment: $1,400 - Monthly ROI: ($12,000 - $1,400) / $1,400 × 100 = 757% ``` #### Cost Optimization Strategies **Strategy 1: Task Prioritization** Focus AI use on highest-value tasks: High ROI Tasks (prioritize): * Large-scale document review * Repetitive drafting (discovery requests, standard letters) * Initial research on novel issues * Document organization and indexing Lower ROI Tasks (use sparingly): * Final brief polishing * Short email responses * Tasks requiring extensive verification * Highly nuanced judgment calls **Strategy 2: Batch Processing** Process similar tasks together to maximize efficiency: ``` Instead of: 10 separate contract reviews at 2 hours each = 20 hours Do: Batch review of 10 contracts together = 8 hours Savings: 12 hours (60%) ``` **Strategy 3: Template Development** Invest time upfront to create reusable prompt templates: ``` Initial investment: 4 hours to develop comprehensive discovery request template Per-use time: 15 minutes to customize template Number of uses per month: 10 Time saved per use: 2 hours Monthly savings: 20 hours (minus 0.25 hours per use = 17.5 net hours) ``` ### Team Training and Firm-Wide Implementation Successfully implementing AI requires training and buy-in across your team. #### Phased Implementation Approach **Phase 1: Pilot Program (Months 1-3)** **Goals**: * Test AI on limited tasks * Develop firm-specific prompts * Establish verification protocols * Measure results **Action Steps**: 1. Select 2-3 pilot tasks (e.g., document review on one case, research for one practice area) 2. Train 2-3 team members as AI champions 3. Document successes and challenges 4. Develop initial prompt library 5. Create verification checklists 6. Measure time/cost savings **Phase 2: Controlled Expansion (Months 4-6)** **Goals**: * Expand to more tasks and team members * Refine protocols based on pilot results * Build comprehensive prompt library * Establish quality metrics **Action Steps**: 1. Train additional team members 2. Expand to additional practice areas 3. Develop firm-wide policies 4. Create internal AI use guidelines 5. Establish regular quality audits 6. Document ROI **Phase 3: Full Integration (Months 7-12)** **Goals**: * AI as standard tool across firm * Continuous improvement processes * Advanced workflow development * Industry leadership **Action Steps**: 1. Mandatory AI training for all attorneys and staff 2. Integration with existing practice management systems 3. Advanced prompt engineering training 4. Regular lunch-and-learn sessions 5. Continuous monitoring and optimization 6. External communication about firm's AI capabilities #### Training Program Template **Module 1: AI Fundamentals (2 hours)** * What is AI and how does it work? * Capabilities and limitations * Ethical considerations * Firm policies and guidelines **Module 2: Basic Prompting (3 hours)** * Three Golden Rules * C.A.S.E. Framework * Prompt Sandwich structure * Hands-on exercises **Module 3: Task-Specific Applications (4 hours)** * Discovery and document review * Legal research * Document drafting * Trial preparation * Practice with real (redacted) examples **Module 4: Quality Control (2 hours)** * Verification requirements * MARP protocol * Common errors and how to catch them * Documentation requirements **Module 5: Advanced Techniques (3 hours)** * Prompt chaining * Multi-step workflows * Custom template development * Troubleshooting and refinement **Total Training Time: 14 hours** (can be delivered over 2-3 weeks) #### Creating Your Firm's AI Policy Every firm should have a written AI policy. Here's a template structure: ``` [FIRM NAME] ARTIFICIAL INTELLIGENCE USE POLICY I. PURPOSE AND SCOPE This policy governs the use of artificial intelligence tools by all attorneys, paralegals, and staff at [Firm Name]. II. APPROVED PLATFORMS The following AI platforms are approved for firm use: - [Platform 1]: For [specific uses] - [Platform 2]: For [specific uses] Unapproved platforms may not be used without prior authorization from [Technology Committee/Managing Partner]. III. CONFIDENTIALITY REQUIREMENTS - Never input client names, case details, or confidential information into public AI platforms - Use only firm-approved secure platforms for matters involving client data - When in doubt, redact and anonymize before using AI - Document all AI use involving client information IV. VERIFICATION REQUIREMENTS All AI-generated content must be verified according to MARP: - Citation validation: [Responsible party and process] - Factual grounding: [Responsible party and process] - Output labeling: [Required labels and sign-offs] V. DISCLOSURE REQUIREMENTS - Check local court rules before filing AI-assisted work - When disclosure is required, use firm's standard certification - Document AI use in matter files VI. QUALITY CONTROL - All AI-assisted work must be reviewed by supervising attorney - Random quality audits will be conducted quarterly - Errors must be reported to [Designated Person/Committee] VII. TRAINING REQUIREMENTS - All attorneys and paralegals must complete AI training within 90 days of hire - Annual refresher training required - Advanced training available for AI champions VIII. PROHIBITED USES - Submitting AI-generated content without verification - Using AI for final decision-making without human oversight - Inputting privileged information into unapproved platforms - Relying on AI citations without independent verification IX. COMPLIANCE AND DISCIPLINE Violations of this policy may result in disciplinary action up to and including termination. X. QUESTIONS AND UPDATES Questions: Contact [Name/Committee] This policy will be reviewed and updated annually. Effective Date: [Date] Last Updated: [Date] ``` ### Chapter Summary Building effective AI workflows requires: * **Strategic Integration**: Map AI use to each litigation phase * **Multi-Step Processes**: Use AI at different stages with outputs feeding forward * **Prompt Chaining**: Break complex tasks into sequential, building prompts * **Quality Assurance**: Implement triple-check systems and ongoing monitoring * **Cost Management**: Track ROI and optimize AI use for maximum benefit * **Team Training**: Phased implementation with comprehensive training programs * **Firm Policies**: Clear written policies governing AI use Key takeaways: * AI is most effective when integrated into systematic workflows * Quality control is essential at every stage * Cost-benefit analysis should guide AI adoption decisions * Training and buy-in are critical for successful implementation * Continuous improvement through monitoring and refinement With these workflows in place, you're ready to transform AI from an experimental tool into a reliable component of your legal practice. *** *In Chapter 6, we'll provide comprehensive resources including an AI platform directory, prompt template library, research papers, and quick reference guides to support your ongoing AI journey.* # 6. Resources and Tools This final chapter provides ongoing reference materials to support your AI journey. As AI technology evolves rapidly, treat this as a starting point and regularly check for updates from AI providers and legal tech resources. ### AI Platform Directory #### General-Purpose AI Platforms These platforms can be used for a wide variety of legal tasks when properly prompted and with appropriate confidentiality safeguards. **ChatGPT (OpenAI)** * **Best For**: Drafting, research, document analysis, creative problem-solving * **Strengths**: Strong reasoning capabilities, excellent at following complex instructions, good for iterative refinement * **Limitations**: Knowledge cutoff (check current date), no real-time web access in base version, potential hallucination * **Confidentiality**: Use API with zero-retention settings for client data; never use free version for confidential information * **Cost**: Free tier available; Plus subscription for enhanced features; API usage billed per token * **Website**: chat.openai.com **Claude (Anthropic)** * **Best For**: Long-form analysis, document review, complex legal reasoning, detailed drafting * **Strengths**: Large context window (handles long documents), strong instruction-following, excellent writing quality, transparent about limitations * **Limitations**: Knowledge cutoff (check current date), conservative responses may require prompting for detailed analysis * **Confidentiality**: Enterprise tier with contractual protections available; API with data retention controls * **Cost**: Free tier available; Pro subscription; Enterprise plans for organizations * **Website**: claude.ai **Google Gemini** * **Best For**: Research, document processing, multimodal tasks (images, PDFs) * **Strengths**: Large context window, integrated with Google services, handles multiple file types * **Limitations**: Rapidly evolving platform; features vary by version * **Confidentiality**: Enterprise options available; check data handling policies carefully * **Cost**: Free tier available; Advanced subscriptions for enhanced features * **Website**: gemini.google.com **Microsoft Copilot (Powered by GPT-4)** * **Best For**: Integration with Microsoft Office suite, web-enabled research * **Strengths**: Web access for current information, integrates with Word/Excel/PowerPoint, available in Edge browser * **Limitations**: Tied to Microsoft ecosystem; data handling varies by subscription type * **Confidentiality**: Microsoft 365 Copilot for business has enterprise data protection * **Cost**: Free in Edge browser; Microsoft 365 Copilot requires enterprise subscription * **Website**: copilot.microsoft.com #### Legal-Specific AI Platforms These platforms are designed specifically for legal work with built-in safeguards and legal training. **Harvey AI** * **Best For**: Large law firms, enterprise legal departments, sophisticated legal analysis * **Strengths**: Trained on legal documents, integrated workflows, compliance features, customizable for specific firms * **Use Cases**: Contract analysis, due diligence, legal research, memo drafting, regulatory compliance * **Confidentiality**: Enterprise-grade security, designed for attorney-client privilege protection * **Cost**: Enterprise pricing (contact for quote) * **Website**: harvey.ai **Thomson Reuters CoCounsel** * **Best For**: Legal research, document review, integrated with Westlaw * **Strengths**: Access to Westlaw database, Practical Law integration, citation checking, multiple AI models * **Use Cases**: Legal research, deposition preparation, contract analysis, timeline creation * **Confidentiality**: Built for law firm use with appropriate data protections * **Cost**: Subscription-based (often bundled with Westlaw) * **Website**: thomsonreuters.com/cocounsel **Lexis+ AI** * **Best For**: Legal research integrated with LexisNexis database * **Strengths**: Direct access to Lexis legal database, Shepard's citations, practice area-specific features * **Use Cases**: Case law research, statute analysis, brief drafting assistance * **Confidentiality**: Designed for legal professional use with data protection * **Cost**: Subscription-based (typically bundled with Lexis Advance) * **Website**: lexisnexis.com/lexis-plus-ai **Casetext (CoCounsel)** * **Best For**: Small to mid-size firms, legal research and document review * **Strengths**: User-friendly interface, strong document review capabilities, cost-effective * **Use Cases**: Legal research, contract review, deposition summaries, legal memo drafting * **Confidentiality**: Attorney-client privilege protections built-in * **Cost**: Subscription-based with various tiers * **Website**: casetext.com #### Document Review and eDiscovery Platforms **Relativity** * **Best For**: Large-scale eDiscovery projects * **Features**: AI-powered document review, predictive coding (TAR), analytics * **Website**: relativity.com **Everlaw** * **Best For**: Litigation teams, document review * **Features**: AI-assisted review, story builder, deposition analytics * **Website**: everlaw\.com **Logikcull** * **Best For**: Small to mid-size matters, affordable eDiscovery * **Features**: Automated document processing, instant insights * **Website**: logikcull.com #### Legal Research Enhancement Tools **ROSS Intelligence** (Note: Check current status) * Legal research AI focusing on case law analysis * Natural language legal queries **Fastcase** * Legal research with AI-powered features * Bad law bot for citation validation ### Prompt Template Library This library provides starting templates for common legal tasks. Customize these for your specific needs following the C.A.S.E. Framework. #### Discovery Templates **Template 1: Deposition Summary** ``` **INSTRUCTIONS** Act as a litigation paralegal preparing deposition summaries for trial counsel. Focus on [specific issues relevant to case theory]. **CONTEXT** Case type: [e.g., Products liability, employment discrimination] This deposition: [Witness name and role] Our case theory: [Brief statement] Opponent's position: [Brief statement] **INPUT** Summarize the attached deposition transcript focusing on: 1. [Key topic area 1] 2. [Key topic area 2] 3. [Key topic area 3] **OUTPUT** Format as: - KEY ADMISSIONS (supporting our case) - INCONSISTENCIES (with other evidence) - DEFENSE-FAVORABLE TESTIMONY - CREDIBILITY ISSUES - FOLLOW-UP TOPICS Include specific page:line citations for all quotes. ``` **Template 2: Privilege Log Review** ``` **INSTRUCTIONS** Act as a privilege review specialist. Flag potentially privileged documents for attorney review. Be conservative - flag anything questionable. **CONTEXT** Case: [Description] Known attorney domains: [List] Review period: [Date range] **INPUT** Review attached documents and identify: 1. Documents with attorney involvement 2. Legal advice or work product 3. Litigation-related communications **OUTPUT** Create three lists: 1. HIGH PRIORITY - Strong privilege indicators 2. MEDIUM PRIORITY - Possible privilege 3. LOW PRIORITY - Marginal indicators Include: Doc ID, Date, From, To, Subject, Privilege Type, Basis ``` **Template 3: Exhibit Organization** ``` **INSTRUCTIONS** Act as a trial preparation specialist organizing trial exhibits. **CONTEXT** Trial type: [Description] Number of exhibits: [X] Trial date: [Date] **INPUT** For each exhibit, extract: 1. Exhibit number 2. Document type 3. Date 4. Author/source 5. Category (from provided list) 6. One-sentence summary 7. Related exhibits **OUTPUT** Format as structured table plus: - Cross-reference notes - Chronological clusters - Trial presentation recommendations ``` #### Legal Research Templates **Template 4: Multi-Jurisdictional Comparison** ``` **INSTRUCTIONS** You are a legal researcher analyzing multi-jurisdictional law. Provide comprehensive analysis with citations to primary sources. If uncertain about current law, state: "This requires verification in Westlaw/Lexis." **CONTEXT** Client situation: [Description] Jurisdictions: [List states/countries] **INPUT** Compare [legal issue] across jurisdictions. For each jurisdiction analyze: 1. General rule 2. Statutory framework (specific citations) 3. Key elements 4. Limitations or requirements 5. Recent changes (last 5 years) **OUTPUT** - Executive summary - Jurisdiction-by-jurisdiction analysis - Comparative matrix (table) - Strategic recommendations - Verification notes ``` **Template 5: Case Law Synthesis** ``` **INSTRUCTIONS** You are a senior associate preparing research for a motion. Analysis should be thorough and cite-checked. For every case cited: 1. Provide full Bluebook citation 2. Include parenthetical explaining holding 3. Note procedural posture 4. Flag if binding or persuasive If uncertain about citation, state: "VERIFY: [explanation]" **CONTEXT** Case: [Description] Motion: [Type] Jurisdiction: [Specific court] **INPUT** Research [specific legal issue] focusing on: 1. [Element 1] 2. [Element 2] 3. [Element 3] Identify: - Leading cases establishing standard - Recent cases (last 10 years) - Factually similar cases - Contrary authority **OUTPUT** Format as research memorandum: - Issue Presented - Brief Answer - Applicable Legal Standard - Analysis (organized by sub-issues) - Contrary Authority - Conclusion - Verification Checklist ``` #### Drafting Templates **Template 6: Discovery Requests** ``` **INSTRUCTIONS** You are a litigation associate drafting discovery requests. Make them specific, targeted, and difficult to object to on vagueness grounds. Guidelines: - Each request seeks specific, identifiable information - Define ambiguous terms - Use time limitations - Avoid compound requests - Include clear response instructions **CONTEXT** Case: [Description] We represent: [Plaintiff/Defendant] Key issues: [List] Discovery goals: [Objectives] **INPUT** Draft [X] interrogatories and [Y] requests for production. Interrogatories should address: 1. [Topic area 1] 2. [Topic area 2] 3. [Topic area 3] Requests should seek: 1. [Document category 1] 2. [Document category 2] 3. [Document category 3] **OUTPUT** Format as formal discovery requests with: - Caption - Instructions - Definitions - Numbered requests - Attestation ``` **Template 7: Demand Letter** ``` **INSTRUCTIONS** You are a litigation attorney drafting a pre-litigation demand letter. Tone: Professional but firm Style: Clear, persuasive, legally sound Avoid: Inflammatory language, unsupported conclusions, empty threats **CONTEXT** Client: [Name and role] Opposing party: [Name] Claim: [Description] Damages: [Itemized] Jurisdiction: [State/federal] **INPUT** Draft demand letter that: 1. Establishes attorney-client relationship 2. Summarizes contract/agreement 3. Details breach/harm with specificity 4. Explains legal basis 5. References applicable law 6. Itemizes damages 7. Makes clear demand 8. Sets reasonable deadline (30 days) 9. Indicates willingness to discuss 10. Preserves litigation option **OUTPUT** Format as formal business letter with: - Law firm letterhead format - Professional tone - Specific demand and deadline - Enclosures list ``` **Template 8: Client Communication** ``` **INSTRUCTIONS** You are a paralegal drafting client communication in plain English (8th grade reading level). Guidelines: - Avoid legal jargon (define if unavoidable) - Use active voice and short sentences - Focus on practical implications - Be honest about risks/uncertainties - End with clear next steps Tone: Professional but warm, informative, honest, reassuring without promises **CONTEXT** Case: [Description] Development: [Recent event] Client background: [Relevant details] **INPUT** Draft email explaining: 1. [Main topic] 2. What it means for the case 3. Our response strategy 4. Timeline and next steps 5. What we need from client (if anything) Keep to 2-3 pages. Use section headings. **OUTPUT** Subject: [Clear description] Sections: - What Happened - What This Means - Our Strategy - Timeline - Next Steps - Questions? ``` #### Trial Preparation Templates **Template 9: Witness Examination Outline** ``` **INSTRUCTIONS** You are a trial attorney preparing witness examination. Develop comprehensive outline that: - Follows logical topic progression - Uses appropriate question types (open-ended vs. leading) - Anticipates objections with responses - Identifies exhibit foundations - Flags potential evasion areas **CONTEXT** Trial type: [Description] Witness: [Name and role] Examination type: [Direct/Cross] Key testimony needed: [List] Challenges: [List] Exhibits to use: [List with numbers] **INPUT** Prepare examination outline for [Witness Name] including: 1. [Topic area 1] 2. [Topic area 2] 3. [Topic area 3] **OUTPUT** Format as detailed outline with: - Section headings by topic - Specific questions - Expected answers - Exhibit references - Anticipated objections and responses - Backup questions - Problem areas and preparation notes ``` **Template 10: Motion in Limine Strategy** ``` **INSTRUCTIONS** You are a trial attorney developing motions in limine strategy. For each potential motion provide: - Specific evidence to exclude/admit - Legal basis with citations - Practical trial impact if granted - Likelihood of success - Strategic importance Be realistic - don't waste time on routinely-denied motions. **CONTEXT** Trial type: [Description] Trial date: [Date] Key evidence issues: [List] **INPUT** Identify and prioritize motions in limine. For each motion: 1. Motion title and relief sought 2. Specific evidence involved 3. Legal standard and governing law 4. Supporting arguments 5. Anticipated opposition 6. Response to opposition 7. Likely ruling 8. Strategic importance (High/Medium/Low) 9. Brief vs. oral argument **OUTPUT** Format as strategy memo with: - Recommended motions (priority order) - Detailed analysis for each - Strategic priorities - Timing recommendations - Hearing strategy ``` ### Further Reading and Research #### Key Research Papers on Legal AI **"Large Language Models as Tax Attorneys: A Case Study in Legal Capabilities Emergence"** * Authors: John J. Nay, et al. * Summary: Examines LLM performance on tax law questions, demonstrating both capabilities and limitations * Key Finding: Advanced models can approach junior associate performance but require verification * Available at: arxiv.org/abs/2306.07075 **"Legal Prompt Engineering for Multilingual Legal Judgement Prediction"** * Authors: Dietrich Trautmann, Alina Petrova, Frank Schilder * Summary: Explores prompt engineering techniques for legal judgment prediction * Key Finding: Zero-shot legal prompt engineering shows promise but falls short of supervised approaches * Available at: arxiv.org/pdf/2212.02199.pdf #### Recommended Books **"The AI-Powered Attorney: How Artificial Intelligence is Transforming Legal Practice"** * Comprehensive overview of AI applications in law * Practical guidance for implementation **"Prompt Engineering for Legal Professionals"** * Detailed techniques specific to legal applications * Case studies from major law firms **"Ethics and AI in Legal Practice"** * Focus on professional responsibility issues * Analysis of disciplinary cases #### Professional Organizations and Resources **ABA Center for Innovation** * Resources on legal technology and innovation * Ethics opinions on AI use * Website: americanbar.org/groups/centers\_commissions/center-for-innovation **Legal Services Corporation Technology Initiative Grant Program** * Information on AI applications for access to justice * Research and pilot programs * Website: lsc.gov/tig **Stanford Center on the Legal Profession** * Academic research on AI in legal practice * Conferences and publications * Website: law\.stanford.edu/legal-profession **International Legal Technology Association (ILTA)** * Legal technology resources and training * Conference presentations on AI * Website: iltanet.org #### Blogs and Newsletters to Follow **Above the Law - Legal Tech** * Daily updates on legal technology developments * Coverage of AI adoption in law firms **Law Technology Today** * Published by ABA Legal Technology Resource Center * Practical articles on legal tech implementation **Artificial Lawyer** * Focus on AI and automation in legal services * Industry news and analysis **LawSites (Bob Ambrogi)** * Reviews and news about legal technology * Interviews with legal tech innovators #### Online Courses and Training **Coursera: AI for Legal Professionals** * Introduction to AI concepts * Legal-specific applications **LinkedIn Learning: Prompt Engineering Fundamentals** * General prompt engineering skills * Transferable to legal applications **Continuing Legal Education (CLE)** * Many state bars now offer AI-focused CLE courses * Check your state bar website for current offerings ### Quick Reference Guides #### The C.A.S.E. Framework Quick Reference **C - Context** * Define subject matter and background * Specify jurisdiction and court * Identify area of law * Reference input data **A - Audience & Action (Persona)** * Assign AI a specific role * Use strong action verbs (Summarize, Draft, Analyze, etc.) **S - Structure & Style** * Define output format * Specify tone (professional, persuasive, plain language) **E - Ethical and Verification Directives** * Require citations * Request confidence levels * Acknowledge limitations * Build in verification requirements #### Prompt Sandwich Template ``` **INSTRUCTIONS** [Persona, ethical directives, general guidelines] **CONTEXT** [Background, jurisdiction, relevant facts] **INPUT** [Specific task, question, or document to analyze] **OUTPUT** [Format requirements, structure specifications, what to exclude] ``` #### Anti-Hallucination Checklist Before relying on AI output: * [ ] Explicit uncertainty instructions included in prompt * [ ] Source attribution requested * [ ] Confidence levels provided * [ ] Every citation verified in primary sources * [ ] Factual assertions traced to source documents * [ ] Legal principles cross-referenced with authoritative sources * [ ] Second AI model consulted for critical conclusions * [ ] Attorney review completed #### Confidentiality Decision Tree ``` Need to use AI? → YES → Client-specific info involved? → YES → Already public? → NO → Can anonymize effectively? → NO → Firm-approved secure platform available? → NO → DO NOT USE AI → YES → Use secure platform only → YES → Anonymize thoroughly, use public AI → YES → May use public AI (exercise caution) → NO → May use public AI (follow verification protocols) ``` #### Verification Protocol Checklist **Citation Validation** * [ ] Case exists in Westlaw/Lexis * [ ] Citation format correct * [ ] Shepardized/KeyCited for validity * [ ] Holding accurately stated * [ ] Procedural posture correct * [ ] Binding vs. persuasive authority confirmed **Factual Grounding** * [ ] Every fact traced to source * [ ] Page/line citations accurate * [ ] Quotations match source exactly * [ ] Context not distorted * [ ] Dates/names/numbers verified **Professional Review** * [ ] Attorney review completed * [ ] Professional judgment applied * [ ] Client-specific application verified * [ ] Appropriate qualifications added * [ ] Quality meets professional standards #### Time Savings Estimation Tool Use this formula to estimate time savings for a specific task: ``` Traditional Time: [X hours] AI-Assisted Time: [Y hours base + Z hours verification] Time Saved: [X - (Y + Z)] hours Value Saved: [Time Saved × Billing Rate] AI Cost: [Platform cost allocated to this task] Net Benefit: [Value Saved - AI Cost] ``` #### Common Prompting Mistakes and Solutions | Mistake | Problem | Solution | | -------------------------- | ------------------------------ | ---------------------------------------- | | Too vague | Generic, unhelpful output | Add specificity using C.A.S.E. Framework | | No context | AI makes incorrect assumptions | Provide background and jurisdiction | | Missing structure | Inconsistent formatting | Specify exact output format | | No verification directives | Increased hallucination risk | Build in uncertainty instructions | | Too complex at once | Overwhelming output | Break into prompt chain | | Confidential info | Ethical violation | Anonymize or use secure platform | ### Staying Current with AI Developments #### Recommended Practices **Weekly** * Scan legal tech news sites for major AI developments * Review any new court rules or ethics opinions on AI use **Monthly** * Test new features from your primary AI platforms * Update your prompt library with successful new templates * Review quality metrics from AI-assisted work **Quarterly** * Attend a CLE or webinar on legal AI * Evaluate new AI tools that have entered the market * Assess ROI and adjust AI integration strategy **Annually** * Comprehensive review of firm AI policy * Team training refresher on latest best practices * Strategic planning for next-generation AI capabilities #### Key Indicators to Watch **Technology Indicators** * New AI model releases from major providers * Expanded context windows (ability to process longer documents) * Improved accuracy rates and reduced hallucination * New multimodal capabilities (voice, video, image processing) **Legal Industry Indicators** * Law firm AI adoption rates * New legal-specific AI platforms * Bar association ethics opinions and guidance * Court rules on AI disclosure **Regulatory Indicators** * Proposed legislation on AI in professional services * Data privacy regulations affecting AI use * Professional liability insurance coverage for AI-assisted work ### Final Thoughts: Your AI Journey You've now completed a comprehensive journey through legal prompt engineering—from understanding the fundamentals to implementing sophisticated workflows. Here are key principles to carry forward: **1. Start Small, Think Big** Begin with one or two high-value tasks where AI can make an immediate impact. As you gain confidence and develop your skills, expand to more complex applications. **2. Always Verify** No matter how sophisticated AI becomes, your professional obligation to verify remains constant. Build verification into every workflow. **3. Protect Confidentiality** Client trust is your most valuable asset. Never compromise it for convenience or efficiency gains. **4. Stay Curious** AI technology is evolving rapidly. The lawyers who thrive will be those who stay curious, experiment thoughtfully, and continuously learn. **5. Maintain Professional Judgment** AI is a tool that amplifies your expertise—it doesn't replace your judgment, ethics, or professional responsibility. **6. Share Knowledge** As you develop effective prompts and workflows, share them with colleagues. The legal profession advances when we learn from each other. **7. Focus on Client Service** Use AI to deliver better, faster, more cost-effective service to your clients. Let improved client outcomes be your measure of success. ### Connect and Continue Learning The field of legal AI is collaborative and rapidly evolving. Consider joining online communities where legal professionals share AI experiences: * LinkedIn groups focused on legal technology * State bar technology sections * Legal tech conferences and webinars * Firm-specific user groups for legal AI platforms ### Acknowledgments This guide builds on the pioneering work of legal professionals, AI researchers, and legal tech innovators who are shaping the future of legal practice. Special recognition to: * The attorneys who learned from early AI mistakes and shared those lessons * Legal tech companies developing secure, ethical AI tools for lawyers * Bar associations providing ethics guidance on AI use * Researchers studying AI applications in legal work ### Guide Updates This guide reflects the state of AI technology and legal practice as of its publication date. For updates: * Check the accompanying website for revised chapters and new templates * Subscribe to the newsletter for significant developments * Follow recommended blogs and resources listed in this chapter *** **You now have the knowledge, tools, and frameworks to integrate AI effectively and ethically into your legal practice. The future of law belongs to professionals who combine deep legal expertise with smart technology use. Go forth and practice law more effectively, efficiently, and ethically with AI as your assistant.** **Good luck on your AI journey!** # 7. Research Corner {% hint style="info" %} Covering the latest advancements in AI, breaking down cutting edge research, and exploring the boundaries of applying artificial intelligence within the field of law. {% endhint %} # 6.1. LLMs as Tax Attorneys In this section we go over the following research paper: > ### Large Language Models as Tax Attorneys: A Case Study in Legal Capabilities Emergence > > **By:** John J. Nay, David Karamardian, Sarah B. Lawsky, Wenting Tao, Meghana Bhat, Raghav Jain, Aaron Travis Lee, Jonathan H. Choi, Jungo Kasai > > **Available at:** ### Introduction The authors of this paper conducted a study to determine whether expert tax attorneys could potentially be replaced with currently available AI models. Tax law was chosen as the subject of this study due to its intricate structure and the necessity for logical reasoning and mathematical skills in its application. Additionally, the legal authority in tax law is principally concentrated in two sources: the Treasury Regulations under the CFR and Title 26 of the U.S. Code (also called the Internal Revenue Code) making it logistically easier to include correct legal texts for reference. Another motivating factor for choosing tax law is the fact that it's deeply intertwined with the real-world economic lives of citizens and companies, making the implications of this study highly relevant and far-reaching. ### Hypothesis The authors hypothesize that LLMs, particularly when combined with prompting enhancements and the correct legal texts, can perform at high levels of accuracy but not yet at expert tax lawyer levels. They also suggest that as LLMs continue to advance, their ability to reason about law autonomously could have significant implications for the legal profession and AI governance. You can get a sense for how this study was structured in the diagram below.
### Evaluation (Prompt Engineering) The authors employed a variety of prompt engineering techniques to enhance the performance of Large Language Models (LLMs) in the context of tax law. These techniques included: 1. **Chain-of-Thought (CoT) Prompting**: This technique involves asking the LLM to think through its response step-by-step. The idea is to encourage the model to generate more reasoned and thoughtful responses. However, the results showed that CoT prompting did not consistently improve results for all models and retrieval methods. It did, however, boost the performance of GPT-4, suggesting that an LLM might need to have a certain capability level to exhibit improved performance through additional reasoning. 2. **Few-Shot Prompting**: In this approach, the LLM is provided with a set of three other question-answer pair examples, along with the question being asked. This is designed to give the model a context and a pattern to follow when generating its own response. The authors found that few-shot prompting significantly improved results for GPT-4 and was less consistently useful for weaker models. 3. **Self-Reflection and Self-Refinement Techniques**: These advanced techniques involve prompting the LLM with its own answers and the relevant context, and then asking it to identify any ambiguities in the question or to doubt its current answer. The response can then be used to conduct further retrieval augmented generation. While the paper does not provide specific results for these techniques, they are identified as prime candidates for increasing performance. 4. **Document Retrieval**: The authors experimented with different retrieval methods, each with its own prompt template that provides different supporting context to the LLM. They found that providing the LLM with more legal text and more relevant legal text weakly increased accuracy for most models. Overall, the results indicated that the effectiveness of these prompt engineering techniques varied depending on the specific LLM and the context. However, they all contributed to enhancing the LLM's ability to reason about tax law and generate accurate responses. ### Evaluation (Models) The paper evaluated the performance of four increasingly advanced Large Language Models (LLMs) released by OpenAI over the past three years. The findings for each model are as follows: 1. **GPT-4**: This was the most advanced model evaluated in the study. The authors found that GPT-4 benefited significantly from the Chain-of-Thought (CoT) prompting technique, which asks the LLM to think through its response step-by-step. This suggests that an LLM might need to have a certain capability level to exhibit improved performance through additional reasoning. GPT-4 also showed significant improvement with few-shot prompting, where a set of three other question-answer pair examples are provided to the LLM, along with the question being asked. Furthermore, GPT-4 showed a clear performance boost when fed with the "gold truth" legal documents, rather than performing similarity search to extract the relevant documents from a vector database. 2. **GPT-3.5**: This model was trained with supervised fine-tuning instead of reinforcement learning from human feedback (RLHF). The results for GPT-3.5 were less consistent than for GPT-4. Few-shot prompting was less useful for this model, and the benefits of CoT prompting were not as pronounced. 3. **GPT-3 (davinci)**: This is the "most capable" GPT-3 model according to OpenAI. The performance of GPT-3 was consistently outperformed by the newer models, indicating the advancements in LLM technology over time. The benefits of advanced prompting techniques were also less pronounced for this model. 4. **GPT-3 (text-davinci-002)**: This is an earlier version of GPT-3.5 that is "trained with supervised fine-tuning instead of RLHF". The performance of this model was similar to that of GPT-3 (davinci), and it was consistently outperformed by the newer models. Overall, the study found that the primary experimental factor causing consistent increases in accuracy was the underlying LLM being used. Newer models consistently outperformed older models, indicating the rapid advancements in LLM technology. ### Results {% hint style="success" %} Senior tax attorneys need not worry about job security for the time being. However, junior tax associates should begin leveraging ChatGPT, or they run the risk of being left behind. {% endhint %} The results of the study provided several key insights into the capabilities of Large Language Models (LLMs) in the context of tax law. Firstly, the study found that the effectiveness of the Chain-of-Thought (CoT) and few-shot prompting techniques varied depending on the specific LLM and the context. CoT prompting, which encourages the LLM to think through its response step-by-step, boosted the performance of the most advanced model, GPT-4, but did not consistently improve results for all models and retrieval methods. This suggests that an LLM might need to have a certain level of capability to benefit from additional reasoning. Few-shot prompting, which provides the LLM with a set of three other question-answer pair examples along with the question being asked, significantly improved results for GPT-4 but was less consistently useful for weaker models. Secondly, the study found that providing the LLM with more legal text and more relevant legal text weakly increased accuracy for most models. This indicates that the quality and relevance of the legal texts used in the prompting process can influence the LLM's ability to generate accurate responses. Finally, and perhaps most importantly, the study found that the primary experimental factor causing consistent increases in accuracy was the underlying LLM being used. Newer models consistently outperformed older models, demonstrating the rapid advancements in LLM technology and their increasing ability to reason about complex subjects like tax law.

Results visualized

These results highlight the potential of LLMs in the legal field, but also underscore the importance of ongoing research and development to further enhance their capabilities. The findings suggest that as LLMs continue to advance, they could play an increasingly significant role in legal services, potentially improving efficiency, reducing costs, and making legal advice more accessible. # 6.2. Prompt Engineering for Legal Judgement Prediction > ### Legal Prompt Engineering for Multilingual Legal Judgement Prediction > > **By:** Dietrich Trautmann, Alina Petrova, Frank Schilder > > **Available at:** ### Introduction Researchers from Thomson Reuters led by Dietrich Trautmann introduce the concept of \ **Legal Prompt Engineering (LPE)**, a process designed to guide and assist LLMs in possibly performing various legal tasks. Their research focuses on Legal Judgement Prediction (LJP), predicting the outcome of a legal case based on the given legal facts, evidence, precedents, and other relevant information. ### Hypothesis Can LLMs be used to automate the prediction of court decisions? More specifically, can legal prompt engineering guide LLMs to effectively perform the LJP task in a zero-shot manner? As a reminder, zero-shot prompting is the most basic form of prompting, and sadly the most common. A zero-shot prompt simply provides a task (ie. ask a question) to the model, nothing more. Earlier in this guide we discussed several prompting strategies like few-example prompting which significantly improves the results of ChatGPT and other LLMs with relatively little effort. **Why did the authors only use zero-shot prompting?** To learn whether or not the implicit (general) knowledge of a LLM translated into a foundational understanding of law. The only additional context provided within the prompts tested were the case texts from the European Court of Human Rights and the Federal Supreme Court of Switzerland. Ultimately, the prompt stack used throughout the experiment is illustrated below.

Legal Prompt Stack

Using zero-shot prompting is the most effective way in testing whether a generic LLM, one that is not further trained or fine-tuned is able to perform legal reasoning. More advanced prompting techniques would greatly skew the results and defeat the purpose of this study. ### Method of Evaluation The authors used *discrete* and *manual* legal prompt engineering. It's a process where they created and evaluated human-readable prompts to classify legal judgments into two categories, a yes/no task. Here's a summary of the process they used: 1. First, they tried using a long legal document as the only input for the language model. The language model tried to continue the document, but the results were not helpful an predicting a guilty or not-guilty prediction. 2. Then, they added a question after the document that reformulated the task. This improved the model's output, but it was still not effective in many cases. Instead of giving a yes/no answer, the model continued with a list of other questions. 3. To improve the model's output, they added the indicators "Question:" and "Answer:". However, the model still gave "free-form" responses, which were difficult to classify into "yes" or "no". 4. They then included answer options "A, Yes" and "B, No" to guide the model's responses. 5. Finally, they used a special indicator to separate the document from the prompt.
{% hint style="info" %} A friendly reminder... **"PROMPT ENGINEERING IS AN ITERATIVE PROCESS."**\ This is a great example of real-world iterative prompt engineering :relaxed:. {% endhint %} ### Results The results reveal that zero-shot LPE performs better than baseline approaches, demonstrating that the transfer to the legal domain is possible for general-purpose LLMs. However, it still falls short compared to the current state-of-the-art supervised approaches. Despite the limitations, the study underscores the potential of LPE in the legal field and its applicability in a multilingual context. ### Why is this research important? AI models can help legal professionals in their decision-making processes, facilitate legal research, and potentially improve the efficiency of legal proceedings. The LJP task being evaluated is challenging due to the complexity of legal language, the need for logical reasoning, and the often extensive length of legal documents. Being able to better understand how LLMs handle complex tasks within specific domains will provide valuable information that can be used to develop strategies for implementing AI to operate effectively throughout entire industries not just limited to law. # Pro Tips! Your one stop shop for quick tips, hacks, and free tools that can transform your law firm. {% content-ref url="/pages/pUmlH9O7HlsLm6BR3Rld" %} [Pro Tip #1](/appendix/pro-tips/pro-tip-1) {% endcontent-ref %} # Pro Tip #1 {% hint style="info" %} TL;DR - Download Microsoft Edge and use AI-powered Bing Chat that is able to provide up to date responses since it has access to data on the internet. {% endhint %} Are you tired of ChatGPT responding with "As an AI language model, I have a knowledge cutoff date because my training data only goes up until September 2021"? You can get around this by upgrading your account to the Plus tier. Or, you can get the same benefits for free by using [Microsoft Edge](https://www.microsoft.com/en-us/edge). Edge is a modern web browser that not only ships with the same bells and whistles included in Chrome and Safari, but they also have AI-powered Bing Chat that functions very similarly to ChatGPT with the addition of being able access data on the internet.

Bing Chat is a free alternative to ChatGPT that can
access the internet to generate responses.

{% embed url="" %} # References Interested in learning more. Take a look at the list of sources, research, and tools we've compiled to deep dive into LLMs, generative AI, and more. ## Research Papers * * * ## Tools * * ## Sources * * * # Change Log All notable changes to this project will be documented in this file. The format is based on [Keep a Changelog](http://keepachangelog.com/) and this project adheres to [Semantic Versioning](http://semver.org/). *** ## 1.2.0 - 2023-07-27 **Added:** * Chapter 6 - This chapter will include breakdowns of research papers covering the latest advancements in artificial intelligence. This chapter will prioritize research being done at the intersection of AI and law. However, absent of extensive research in this niche area, we'll be covering more technical research papers, in which we'll include a legal spin on our writings with plenty of real-world examples. * Section 6.1 - Can current LLMs like ChatGPT outperform senior tax attorneys?! That is the question Stanford researcher John J. Nay and team set out to answer. Tl;DR - AI models can currently perform at a level comparable to a junior tax associate. However, each subsequent version of advanced LLMs perform incrementally better. The jobs of tax attorneys are safe, for now 😜. * Section 6.2 - The authors' main focus is to use LPE with LLMs on lengthy legal documents for Legal Judgement Prediction (LJP). An interesting approach the researchers took was to use general LLMs with **NO** prompt engineering. Hence, they wanted to figure out whether the general knowledge of LLMs would be applicable to the specialized subject of law. ## 1.1.0 - 2023-07-20 **Added:** * [Section 4.3](/4.-ethical-guardrails-and-professional-responsibility/4.3.-chatgpt-plugins) - We cover ChatGPT plugins in detail starting with examples that detail how you can ChatGPT generate responses from data included in PDF documents you upload, and how you can use Plugins to connect to external data sources like the internet. Make sure to interact with our interactive demos to easy learn step-by-step. * [Section 4.4](/4.-ethical-guardrails-and-professional-responsibility/4.4.-chatgpt-code-interpreter) - Code Interpreter is an exciting addition to ChatGPT that lets you accomplish more complex tasks like generating PDFs, solving complex maths, data analysis, and more. * [Pro Tips!](/appendix/pro-tips) - We are going to start sharing some nifty hacks, tricks, and free tools we use to be more effective when using ChatGPT. Caring is sharing :innocent:. * [Change Log](/appendix/change-log) ## 1.0.0 - 2023-07-15 #### :confetti\_ball: **Initial release of LegalPromptGuide.com** :confetti\_ball: **Added:** * Chapters [1](/1.-introduction-the-power-of-precision-in-prompting), [2](/2.-fundamentals-of-legal-prompt-engineering), [*4*](/4.-ethical-guardrails-and-professional-responsibility), [5](/6.-resources-and-tools) * [Section 4.2](/4.-ethical-guardrails-and-professional-responsibility/4.2.-output-parsers) * [References](/appendix/references)