The Problem
Generic AI responses lack the depth, tone, and perspective you need. A contract review from "an AI" feels different from a review by a senior M&A lawyer with 15 years' experience in England and Wales. You waste time adding context and refinement.
The Solution
Start your prompts with a precise role assignment:
"You are a [specific role with experience details]. Respond as that person would."
The more specific you are about expertise, jurisdiction, and context, the better the output.
Template
You are a [seniority level] [practice area] lawyer with [X years]
experience in [jurisdiction]. You work at [firm type] and specialise
in [specific area].
[Your task]
Respond as that person would - with appropriate depth, tone, and
jurisdiction-specific knowledge.
Before You Rely on This
The examples on this page are made up to show the technique. No test results have been published for it, so try it on your own work and judge it by what you find.
This guide is about how to word a prompt. It does not make a client's document or information safe to put into an AI tool. What you may enter depends on the plan your firm has, its agreement with the vendor and your firm's policy: Claude Projects for Client Matters sets out the checks for one tool. To practise, use a precedent or a published document that contains no client information, and refer to parties by role.
Why It Works
Anthropic's guidance for Claude says: "Setting a role in the system prompt focuses Claude's behavior and tone for your use case." (checked 29 September 2026; the page is listed under "Source"). That guidance is written for developers, and it is one vendor writing about its own models. How far a role changes the substance of an answer has not been tested for this guide, so compare the two versions on your own work.
Example
Generic prompt:
Review this share purchase agreement for risks.
AI response: Generic, shallow analysis. Mentions obvious issues like "review warranties carefully."
With role assignment:
You are a senior corporate lawyer with 15 years of M&A experience in
England & Wales. You work at a Magic Circle firm and specialise in
private equity transactions.
Review this share purchase agreement for risks that would concern a
PE investor. Focus on deal-breakers and material issues, not minor
drafting points.
Respond with the depth and perspective a partner would provide.
AI response: Analysis aimed at a PE investor's concerns, such as management warranties, locked-box mechanism risks, W&I insurance implications and completion accounts disputes. A role changes the register of an answer. It does not make the answer right, so check the substance.
Pro Tips
Be specific about:
- Seniority: "Trainee" vs "Senior Associate" vs "Partner" changes depth
- Practice area: "Employment lawyer" vs "Sports employment lawyer"
- Jurisdiction: "UK" vs "England & Wales" vs "Scotland"
- Firm type: "Magic Circle" vs "high street firm" vs "in-house"
- Specialisation: "Commercial litigation" vs "insurance litigation"
Examples:
You are a newly qualified solicitor in a regional firm, advising
small businesses on standard commercial contracts.
You are a General Counsel at a FTSE 100 company with 20 years of
in-house experience, focused on regulatory compliance and governance.
You are a barrister with 10 years' call specialising in employment
tribunal advocacy, primarily acting for claimants.
Common Mistakes
❌ Too vague: "You are a lawyer." (Tells the AI almost nothing)
❌ Contradictory: "You are a junior trainee with 30 years of experience." (Confusing)
❌ Irrelevant detail: "You are a lawyer who likes tennis and has two dogs." (Doesn't affect output)
✅ Specific and relevant: "You are a partner in employment law with expertise in executive terminations and restrictive covenants."
Quick Reference
Copy and customise:
You are a [seniority] [practice area] lawyer with [X years] experience
in [jurisdiction]. You [work at/specialise in] [context].
[Task]
Respond with the depth, tone, and perspective this role would provide.
Combine with:
- "The Assumption Buster" to prevent jurisdiction guessing
- "Output Format Control" to specify how the response should be structured
- "Step-by-Step Reasoning" for complex analysis
Source
Checked on 29 September 2026.
- Anthropic, Prompting best practices: the section "Give Claude a role"