Write the firm's AI policy, and make the associates read it

1 · Learn the move · Recipe pattern

A policy that says 'use AI responsibly' is a policy nobody can violate. The one that works is a recipe: numbered steps a lawyer or a paralegal follows on a specific afternoon, each with a yes-or-no at the end. Which tools and plans are approved, by name, and the person who keeps that list current. What never goes in, by category, with the de-identification template from this desk as the step before any paste. What always comes out marked: every authority VERIFY, every quotation UNSEEN until opened. The standing-order check before any court filing. The consent step for any use that touches information relating to the representation in a self-learning tool. The billing line. The supervision line: the rules put the supervising lawyer on the hook for the nonlawyer's use as much as their own [M], so the policy names who checks whom. And a signature line, because a policy nobody signed is a memo. The model drafts the recipe from the decisions you have already made; where you have not made one, it leaves the step as [CHECK THE RECORD: the firm's decision] rather than deciding for you, and it never writes that following the policy satisfies any rule.

Draft our firm's AI use policy as a numbered recipe, from these decisions: approved tools and plans [list], who keeps the list [role], what never goes in [categories], the de-identification step [paste our template], the authority rule (VERIFY until opened; quotations UNSEEN), the standing-order check before filing, the consent step and who obtains it, the billing line, the supervision line, the signature line. Every step ends in a yes-or-no a person can answer. Any decision I did not give you is [CHECK THE RECORD: the firm's decision]. Do not write that following the policy satisfies any rule.