Prompt Progression Drill

Purpose

This exercise helps participants understand that good legal prompting is not about finding magic words. It is about turning an underspecified request into a supervised legal research assignment.

Participants start with a weak legal research prompt, diagnose what is missing, and revise it into a stronger version. They add the legal and practical information a research assistant would need: jurisdiction, timeframe, authority scope, task type, relevant facts, constraints, output format, and a plan for what must be verified next.

The point is not to produce one perfect prompt. The point is to build the habit of prompt progression: weak to better to stronger, with each revision making the task clearer, more constrained, and easier to evaluate.

Learning Goals

By the end of this exercise, students should be able to:

1. Identify why a weak legal research prompt is likely to produce vague, overbroad, or unreliable output.
2. Revise a prompt by adding jurisdiction, timeframe, source type, court level, facts, client posture, and task type.
3. Distinguish between prompts that ask for orientation, search strategy, source identification, legal analysis, drafting, or verification.
4. Add constraints that reduce risk, including facts-only instructions, no speculation, source limits, format requirements, and assumptions or uncertainty flags.
5. Build a follow-up verification plan before relying on the AI output.
6. Explain how prompt quality affects the usefulness and reviewability of the output.

Materials

• Three to five weak legal research prompts.
• A prompt progression worksheet with three columns:
  • Weak prompt.
  • Better prompt.
  • Strong prompt.
• A prompt checklist including:
  • Jurisdiction.
  • Timeframe.
  • Court level or authority scope.
  • Source type.
  • Task type.
  • Relevant facts or client posture.
  • Constraints.
  • Output format.
  • Follow-up verification plan.
• Optional: access to an AI tool so participants can test the weak and revised prompts and compare outputs.
• Optional instructor materials: sample weak, better, and strong prompts; model revisions; debrief questions.

Setup

This exercise works best after participants have been introduced to the idea that prompting is a form of supervision.

Before beginning, participants should understand that different legal research tasks require different prompt structures. A prompt asking for general orientation does not need the same level of specificity as a prompt asking for controlling authority in a particular jurisdiction. A prompt asking for a Boolean search string is different from a prompt asking for an issue-spotting checklist or research roadmap.

The instructor should frame the exercise around iteration. The first prompt is rarely the final prompt. A good legal research workflow often begins with a broad orientation prompt, then narrows through follow-up questions, source checks, and verification.

Run of Show

Total, approximately 15–25 minutes

1. Frame the task Explain that participants will revise weak legal research prompts into stronger prompts. The goal is not longer prompts. The goal is clearer prompts that produce outputs a lawyer can evaluate.
2. Diagnose the weak prompt Give participants a weak prompt. Ask them to identify what is missing: jurisdiction, timeframe, source type, authority scope, facts, court level, task type, output format, or verification plan.
3. Build the better prompt Participants revise the prompt by adding the most important missing context. This version should be clear enough to produce a more relevant answer but may still need additional constraints.
4. Build the strong prompt Participants revise again, adding constraints, output format, and a follow-up verification plan. The strong prompt should make clear what the AI may do, what it may not do, and what the lawyer will check afterward.
5. Optional tool test Participants run the weak prompt and the strong prompt in an AI tool, then compare outputs. They should focus on whether the stronger prompt produced better structure, better jurisdictional focus, clearer uncertainty, and a more useful research path.
6. Debrief Discuss what changed between the weak and strong versions. Focus on how the revised prompt made the work easier to supervise and verify.
7. Submission or reflection Participants submit one weak-to-strong prompt progression and a one-sentence explanation of the verification step they would take next.

Student Instructions

Instructor Notes

The teaching value of this exercise comes from showing that prompt revision is really task clarification.

Participants may initially improve weak prompts by adding more words. Push them to add better legal instructions, not just more language. A strong prompt should narrow the task, define the source universe, set boundaries, and make verification easier.

A useful sequence is: Weak, “Find cases about noncompetes.” Better, “Find recent Tennessee cases about enforceability of noncompete agreements against sales employees.” Strong, “Act as a junior associate creating a research roadmap for a supervising attorney. Identify Tennessee cases from 2020 to the present addressing enforceability of noncompete agreements against sales employees. Focus on cases discussing reasonableness of scope, duration, geography, and protectable business interests. Provide a short table with case name, court, year, key facts, holding, and why the case may matter. Do not state that a case is controlling unless you identify the court. Flag any uncertainty and list what I should verify in Westlaw or Lexis before relying on the cases.”

The strong version is not better because it is longer. It is better because it tells the AI what role to play, what law to search, what sources matter, what output format is useful, what claims require caution, and what must be verified.

Common issues to watch for:

• Participants may forget jurisdiction.
• Participants may ask for “the law” without defining authority scope.
• Participants may fail to distinguish between finding sources and analyzing sources.
• Participants may ask AI to provide citations without planning to verify them.
• Participants may omit timeframe even when recency matters.
• Participants may ask for a final answer when they really need a research roadmap.
• Participants may forget to require uncertainty flags or assumptions.

Strong work usually shows:

• A narrowed legal task.
• Clear jurisdiction and timeframe.
• Defined source or authority scope.
• Useful output format.
• Constraints against speculation.
• Instructions to flag uncertainty.
• A concrete verification step.

Adaptation

Here are some suggestions on how to adapt this in other teaching contexts. The adaptations below have been limited to firm trainings and asynchronous suggestions, but there are other possible adaptations for workshops, seminars, and other contexts. Be creative!