Generative AI can do great legal work when we bring our analytical precision to the prompting.
Lawyers are training skeptics. When AI output is inconsistent, concluding that the tool isn’t ready for legal work isn't’ a failure of imagination, its professional judgment at work. The real questions isn’t whether AI is ready for Legal work, it’s whether the instructions we give to the AI are ready. Vague prompts produce vague output - this is not an AI problem. It’s a prompting problem, one attorneys are uniquely positioned to solve.
The same discipline that governs good contract drafting – clarity, context, structure and precision – is exactly what produces strong AI output.Treat it like a junior associate: give it context, coach it toward the right answer, and break complex tasks into steps.Below are five AI prompts tailored to common in-house legal workflows. Each is designed to produce more useful output and to show you how you can adapt it to your own legal needs.
Prompt:
You are the in-house counsel negotiating a vendor agreement.
TASK: Rewrite the limitation of liability clause to reflect a customer-favorable position.
OUTPUT FORMAT: Redlined clause (show additions and deletions) with external facing supporting commentary.
CONTEXT: The company is the customer. Target: 12 months’ fees cap. Exclude indemnity, confidentiality, and data breach from the cap. Vendor is likely to push back on anything above fees paid.
COMMENTARY TONE: Professional and commercially reasonable
CONTENT: [Insert clause]
Why this prompt works
AI can draft effectively, but only when constraints mirror real negotiations.
This prompt:
The output will be usable, not theoretical.
Prompt:
You are a senior commercial counsel building a [agreement type] contract negotiation playbook for an in-house legal team.
TASK: Analyze the attached agreements and extract standard positions, acceptable fallbacks, and known risk thresholds by clause type.
OUTPUT FORMAT: Table with columns: Clause Type, Standard Position, Acceptable Fallback Positions, Risk, Notes.
CONTEXT: B2B SaaS company selling to enterprise customers; priority is deal velocity within acceptable risk tolerance.
CONTENT: [Insert standard agreement and 3–5 executed agreements]
Why this prompt works
Rather than asking AI to generate a playbook from scratch, this approach:
AI performance is accelerated when grounded in tangible examples rather than abstract and generalized requests. Just be sure you are consistent in the context of the examples provided (e.g. same type of agreement) , because NDAs aren't good training data for MSAs, and vice versa.
Prompt:
You are General Counsel preparing a briefing for the CFO.
TASK: Summarize the key legal and financial implications of this agreement.
OUTPUT FORMAT: 5 bullet points, each under 25 words.
AUDIENCE: The audience is financially sophisticated but not legally trained. CONSTRAINTS: Focus on risk exposure, cost implications, and operational constraints.
TONE: Direct, pragmatic, no legal jargon.
CONTENT: [Paste agreement or summary]
Why this prompt works
Legal summaries often miss the mark because they’re written for other lawyers, not business decision-makers.
This prompt forces the model to:
Not every AI task can be effectively completed in a one-shot prompt. For more complex tasks, use a prompt sequence.
CONTEXT FOR ALL PROMPTS IN THIS SEQUENCE: You are supporting in-house counsel at a U.S.-based mid-market SaaS company (customer/licensee). Risk tolerance is moderate. Priority is protecting against uncapped liability and broad indemnities while maintaining deal velocity. The counterparty is a large enterprise vendor unlikely to accept non-standard terms without justification.
Prompt sequence:
Prompt 1:
Draft three versions of a personal data processing clause for a SaaS MSA, ranging from customer-friendly to vendor-friendly.
Prompt 2:
We are the customer. Select the version that best protects us while remaining commercially reasonable for a mid-market deal. Explain your reasoning in two sentences.
Prompt 3:
Revise the selected clause to address GDPR compliance obligations and add audit rights.
Why this prompt works
Treating AI like a thoughtful collaborator that needs context, constraints, and clear objectives, is what unlocks its potential, and with the LinkSquares Agentic Platform you can create a step-by-step prompt sequence:
Breaking work into steps:
For drafting truly novel legal clauses, iteration isn’t optional; it’s the mechanism that drives quality.
Prompt:
You are a senior commercial counsel reviewing a SaaS master services agreement.
TASK: Identify and prioritize legal risks in the attached agreement.
OUTPUT FORMAT: Table with columns: Clause, Risk Level (High/Medium/Low), Why It Matters (operational and commercial impact), Suggested Revision.
CONTEXT: The company is a mid-market SaaS company, U.S.-based. We are reviewing as customer with moderate risk tolerance and need toavoid uncapped liability and broad indemnities.
CONTENT: [Paste contract or clause]
Why this prompt works
Generic instructions like “review this contract” produce generic output.
This version works because it:
The result will be an actionable issue list rather than a high-level summary.
The teams I see getting the most out of AI aren't just writing better prompts – they're making sure those prompts don't live and die in one person's chat history. Legal teams looking to build institutional rigor should consider the following approaches for embedding prompts into their ’ workflows:
Effective prompting isn’t a technical trick, it’s disciplined thinking conveyed via explicit instructions.
Strong prompts consistently:
Prompting is strategic communication. You've been doing this your whole career – AI just gave it a new application.