Asking well · for legal and compliance professionals
Asking well in legal and compliance work
A request such as "review this contract" gives the model no standard to review against. It does not know which party you represent, which clauses matter to your company, or what a departure from your position looks like. It returns a general summary that reads well and decides nothing.
Asking well means putting the questions a senior lawyer would ask into the request. Whose side are we on? Which clauses are in scope? What is our fallback position? What form should the answer take, and what must it never do, such as state a conclusion on enforceability?
Where it comes up
Clause comparison against a playbook
You need a vendor’s limitation of liability clause compared with your standard position. A strong request states which party you are, quotes your preferred language, lists your acceptable fallbacks, and asks for each departure in a table with the clause number.
A regulation summary for a product team
Product needs to know whether a new consumer protection rule affects a feature. The request names the audience, the feature, the sections of the rule in scope, and asks the model to quote the text behind each point.
First-pass redlines on a services agreement
A good request says which clauses to redline, which to leave alone, and how to mark each change. It asks for a short reason beside each edit so a reviewer can accept or reject it quickly.
What it looks like at each band
E1 · Foundational
Gets usable results on familiar tasks with simple requests. Requests name the document and the task but often leave out the party, the standard, and the scope. Improves weak output by rewriting it by hand rather than by fixing the request.
E2 · Proficient
States the party represented, the standard to review against, the clauses in scope, and the output format. Asks the model to cite clause numbers or quote the text for each point. When an answer misses, adds the missing instruction and asks again.
E3 · Distinguished
Splits complex reviews into stages, such as issue spotting first and drafting second. States what the model must not do, such as reach a conclusion on a contested point. Asks the model to identify ambiguities in the instructions before it starts.
E4 · Exceptional
Writes request templates the team uses for recurring matters, such as NDA review or vendor intake, with fields for party, risk position, and jurisdiction. Knows which tasks need one exact request and which need a short exchange. Colleagues adopt these templates because they produce consistent reviews.
Common mistakes
Asking for a review with no standard
A model cannot tell you a clause is unfavorable unless it knows what favorable means to you. Give it your position or your playbook language.
Not saying which party you represent
The same indemnity clause is good for one side and bad for the other. Leave this out and the model either guesses or hedges every point.
Asking for answers without the source text
An answer that does not quote or cite the clause or section is hard to check. Ask for the reference with every point.
A worked example
A first-pass review of a customer’s proposed changes to your standard SaaS subscription agreement.
Review these contract changes and tell me if there are any problems.
The model does not know whose agreement it is, which changes matter, or what counts as a problem. It will list every change with a general comment and no priority.
We are the vendor. The attached redline is the customer’s markup of our standard SaaS agreement. Review only the changes to limitation of liability, indemnity, data protection, and termination. For each change, give the clause number, quote the new text, compare it with our standard text, and rate it as accept, negotiate, or escalate. Use our fallback positions below. Do not give a view on enforceability. Present the result as a table.
The request names the party, the clauses in scope, the comparison standard, the fallbacks, the output format, and a clear limit. A reviewer can check each row against the document.
The full study guide for this skill, for every role, is Asking well. The other five skills for legal and compliance professionals are on the AI skills for legal and compliance professionals page. A one-page summary is in the AI prompt checklist.
See where you stand
The free check grades asking well and the other five skills in about fifteen minutes. Nobody sees the grade but you.
Then build the skills and prove it. The three steps for legal and compliance professionals list the courses and templates that fit the work.