The Problem
You ask AI a question, get a generic answer, then spend time providing context that should have been included from the start. Each round of "actually, here's more context..." wastes time.
The Solution
Front-load all relevant context before asking your question. Structure your prompts:
Context → Question → Constraints
Give the AI the full picture first, then ask what you need.
Template
CONTEXT:
[All relevant background information]
QUESTION:
[What you need]
CONSTRAINTS:
[Any limitations or requirements]
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.
Give the facts the question needs and no more. Figures, dates and descriptions can identify a client even when no name is given.
Why It Works
LLMs generate responses based on the information available at the time of generation. If you drip-feed context across multiple messages, earlier responses are based on incomplete information. Front-loading context lets the AI consider everything from the start.
What Context to Include
For Legal Analysis:
- Jurisdiction: England & Wales / Scotland / Northern Ireland
- Parties: Public company / private individual / startup
- Transaction type: M&A / funding / restructuring
- Your role: Acting for buyer / seller / company / investor
- Stage: Pre-heads of terms / due diligence / post-signing
- Constraints: Budget / timeline / political sensitivities
For Contract Work:
- Contract type: SaaS / employment / IP assignment
- Industry: Tech / finance / manufacturing
- Commercial context: New relationship / renewal / distressed
- Negotiation status: First draft / third round / final push
- Key concerns: Liability / IP / termination / pricing
For Dispute Work:
- Type of dispute: Employment / commercial / IP
- Stage: Pre-action / early litigation / trial prep
- Opponent: Represented / litigant in person / aggressive
- Budget: Limited / uncapped / fixed fee
- Client goals: Win / settle / principle / money
Examples
❌ Bad (No Context)
Can we terminate this agreement?
AI response: Generic answer about termination clauses. Mixes the law of several jurisdictions. Discusses multiple scenarios. Not helpful.
✅ Good (Context Front-Loaded)
CONTEXT:
- We are a UK SaaS company
- 3-year software licence agreement with a customer
- Customer has been in material breach for 4 months (non-payment)
- Contract is governed by English law
- We've sent 2 warning notices (no response)
- Customer is a large corporate (not insolvent)
- We want to exit the relationship, not chase payment
QUESTION:
Can we terminate this agreement immediately, or must we follow a specific process?
CONSTRAINTS:
- Assume we have the contract available if you need to see specific clauses
- Focus on practical steps, not theoretical analysis
- 300 words max
AI response: Advice on termination for material breach under English law, with specific steps to take. Check it against the contract before you act on it.
Pro Tips
Structure your context into categories:
CONTEXT:
Commercial background:
[Business situation, parties, transaction type]
Legal background:
[Jurisdiction, existing agreements, regulatory context]
Current situation:
[What just happened, immediate issue]
Our client:
[Who we represent, their goals, constraints]
QUESTION:
[Your specific question]
Include negative context too:
What we don't know yet:
- Exact amount of customer's losses
- Whether they've already instructed solicitors
What's not relevant:
- Tax treatment (handling separately)
- IP ownership (already resolved)
Common Mistakes
❌ Too much context: 5 paragraphs of background when 3 sentences would do
❌ Irrelevant context: Including information that doesn't affect the answer
❌ Buried question: Putting the question in the middle of context (put it after context, clearly labelled)
✅ Relevant, concise context that gives AI everything it needs and nothing it doesn't
Quick Reference
Standard Template:
CONTEXT:
[Who]: [parties, your role, jurisdiction]
[What]: [transaction/dispute type, current stage]
[Why relevant]: [business context, client goals]
[Constraints]: [budget, timeline, sensitivities]
QUESTION:
[Your specific question, clearly stated]
OUTPUT REQUIREMENTS:
[How you want the answer formatted]
Worked Example
The matter below is made up. It shows the format.
CONTEXT:
Acting for a tech startup (UK company, 30 employees) in a Series A funding round.
We've received a term sheet from a US VC fund. £5m investment for 25% equity.
This is our client's first institutional funding (previously bootstrapped).
The term sheet includes standard liquidation preference, anti-dilution protection,
and board seats for the investor.
Our client's main concerns:
1. Maintaining control over business decisions
2. Protecting against dilution in future rounds
3. Understanding what's market-standard vs aggressive
Our client is sophisticated (tech background) but new to VC fundraising.
QUESTION:
What are the top 3 commercial terms we should focus on negotiating,
and what's a realistic position to take on each?
CONSTRAINTS:
- High-level strategic advice (not full term sheet review)
- Assume this VC is a Tier 1 fund (strong negotiating position)
- 400 words max
- Prioritise by importance
What to look for: advice that uses the context you gave. Check it as you would any AI output.