I recently finished a trial. It was a blur of long nights, last-minute pivots, and, what I would call, "controlled chaos".

Over a few posts, I will share some practical lessons I learned about using AI during trial.

Let's start with something simple I recommend you do before trial, which will pay dividends throughout: building what I call a "Context Document" for AI.

The Problem: AI Has No Idea What Your Case Is About

Every time you start a new conversation with an AI tool, you are talking to someone with complete amnesia about your matter. It does not know your client's name, the underlying dispute, the key witnesses, or the themes you have been developing for months. It does not know that the defendant's CEO gave damaging testimony at his deposition, or that there is a smoking gun email from October 2022 that anchors your entire narrative.

This means that every time you need AI help during trial, you face a choice: either spend five+ minutes explaining the case from scratch, or give a bare-bones prompt and get generic output that requires heavy editing.

Neither option is great.

The Solution: Build Your Context Document Before Trial

A Context Document is essentially a comprehensive briefing memo written specifically for AI consumption. It tells the AI everything it needs to know to help you effectively, formatted in a way that makes the information easy to process and apply.

Think of it as onboarding a new associate to your case, except this associate has perfect recall, works instantly, and will never bill you for the time it takes to get up to speed.

The best time to create this document is during trial prep, when you are already synthesizing your case narrative for opening statements. You have your themes crystallized. You know your key facts. You have your witness list organized. All of that thinking can be captured and repurposed.

How to Create Your Context Document

Here is a process you can try:

Step 1: Gather Your Materials

Pull together:

  • Your opening statement outline or draft
  • Your case narrative or theory of the case memo
  • Key facts and timeline
  • Witness list with brief descriptions
  • Any theme documents or message triangles you have developed

Step 2: Feed It to your Firm-Approved AI with Clear Instructions

Here is a sample prompt you can adapt:

<role>
You are an expert legal writing assistant helping a 
litigation attorney prepare reference materials for trial.
</role>

<task>
Based on the materials I provide below, create a comprehensive 
"AI Context Document" that I can use to quickly brief any AI 
assistant on my case throughout trial.

The document should be written in second person, addressed to 
an AI assistant (e.g., "You are helping counsel for the Plaintiff...").

Include the following sections:

1. CASE OVERVIEW
   - Parties and their roles
   - Nature of the dispute
   - Key claims and defenses
   - Current procedural posture

2. OUR CLIENT'S POSITION
   - Core narrative and theory of the case
   - Key themes (2-3 main themes we are emphasizing)
   - What we need to prove / our strongest arguments

3. KEY FACTS AND TIMELINE
   - Critical dates and events
   - The "story" in chronological order
   - Pivotal documents or evidence

4. KEY PLAYERS
   - Witnesses and their significance
   - Key decision-makers
   - Relationships between parties

5. WHAT TO KNOW WHEN HELPING ME
   - Our client's tone and positioning preferences
   - Sensitivities or topics to handle carefully
   - Common opposing arguments we need to counter
</task>

<materials>
[PASTE YOUR OPENING STATEMENT OUTLINE HERE]

[PASTE YOUR CASE NARRATIVE OR THEORY MEMO HERE]

[PASTE YOUR WITNESS LIST HERE]

[ADD ANY OTHER RELEVANT MATERIALS]
</materials>

<output_requirements>
- Write in clear, direct prose
- Aim for approximately 1,500 to 2,000 words
- Use formatting that will be easy to paste into future prompts
- Focus on information that would help an AI assistant 
  provide better, more case-specific help
</output_requirements>

Step 3: Review and Refine

The AI will generate a first draft. Review it carefully for:

  • Accuracy (did it capture the facts correctly?)
  • Completeness (is anything critical missing?)
  • Tone (does it reflect your actual positioning?)
  • Usability (is it formatted for easy copy/paste?)

Make edits directly, or ask the AI to revise specific sections.

Using Your Context Document During Trial

Once you have your Context Document, using it is simple. Whenever you need AI assistance, start your prompt by pasting the document as background context, then provide the specific materials you want the AI to work with.

Here is a sample you could use for jury selection, for example:

<case_context>
[PASTE YOUR CONTEXT DOCUMENT HERE]
</case_context>

<task>
Jury selection is tomorrow morning. I am pasting the jury 
questionnaire responses below. Using the case context as 
background, help me prepare for voir dire by doing the 
following:

1. JUROR-BY-JUROR ANALYSIS
   For each juror, identify:
   - Any responses that suggest potential bias for or 
     against our client's position
   - Life experiences or professional background that 
     may shape how they view the key issues in this case
   - Red flags that might warrant a strike
   - Positive indicators that suggest alignment with 
     our themes

2. FOLLOW-UP QUESTIONS
   For jurors with concerning or ambiguous responses, 
   draft 2-3 open-ended follow-up questions designed to 
   surface whether the concern is serious enough to 
   justify a cause challenge or peremptory strike. 
   Questions should be conversational and non-leading.

3. PRIORITY RANKING
   Based on your analysis, rank the jurors from most 
   favorable to least favorable for our position, with 
   a brief explanation for each ranking.

4. STRIKE RECOMMENDATIONS
   Identify which jurors you would recommend striking 
   and why, distinguishing between "definite strikes" 
   and "strike if possible."
</task>

<questionnaire_responses>
[PASTE RESPONSES HERE]
</questionnaire_responses>

<output_format>
Organize your analysis so I can quickly reference it 
during voir dire. Use juror numbers as headers and 
keep the follow-up questions in a format I can read 
naturally in the courtroom.
</output_format>

Notice the pattern: the Context Document provides background understanding of your case, while the questionnaire responses are the actual material the AI analyzes. The AI now has everything it needs to give you case-specific, actionable output. No more generic advice. No more wasted time explaining your theory of the case from scratch.

If your firm uses an AI product that does not permit the integration of client-specific information, there is a solution available.

The solution is to create a second, stripped-down version of your Context Document that provides legal context without revealing client identity or strategy.

Here is a prompt to generate it:

<task>
Based on the AI Context Document below, create a 
"Legal Research Context" version that I can use with 
legal research AI tools like Westlaw AI and Lexis AI.

This version should:

1. REMOVE all of the following:
   - Client names (replace with generic descriptors 
     like "Plaintiff" or "the technology company")
   - Specific party names
   - Case name or docket number
   - Strategic positioning or themes
   - Any information that reveals which side we represent

2. KEEP the following:
   - General nature of the dispute
   - Relevant legal claims and defenses
   - Jurisdiction and court type
   - Key legal issues
   - Factual pattern in generic terms

3. FORMAT as a brief paragraph (150-200 words) that can 
   be pasted at the start of a legal research query
</task>

<context_document>
[PASTE YOUR FULL CONTEXT DOCUMENT HERE]
</context_document>

The output will give you something like this:

"This matter involves a breach of contract dispute in federal court (SDNY) between a mid-size technology company and a former business partner. The core claims include breach of a services agreement, breach of fiduciary duty, and misappropriation of trade secrets. Key legal issues include the enforceability of a limitation of liability clause, the application of the business judgment rule to pre-termination conduct, and whether certain communications are admissible under the business records exception to hearsay. The factual pattern involves a multi-year business relationship that deteriorated after one party allegedly began diverting business opportunities and misusing confidential information shared under NDA."

Now when you run a research query in Westlaw AI, you can preface your question with this context, and the research AI will tailor its results to your specific situation without you exposing confidential details.

The Payoff

The attorneys who get the most value from AI during trial will not be the ones who use it the most. They will be the ones who use it most efficiently.

Building your Context Document takes maybe a few minutes during trial prep. That investment pays for itself the first time you need to draft a motion response overnight and can get a solid first draft in three minutes instead of thirty, or an hour.

Trial is too fast and too high-stakes for generic AI output. Building the infrastructure to get case-specific help will get AI results that are actually usable.