AI adoption for managing partners. general counsel. innovation officers. practice group leaders. knowledge teams. risk & compliance. small firms. litigation teams.

It is a reference and a diagnostic, not a vendor. Firm leadership faces a set of decisions — what to adopt, who checks the output, what to charge, what happens to client confidences — and almost everything written about those decisions comes from someone selling a product. The Institute exists to be the part that is not selling anything: ten areas of plain-language reference written for people who decide, plus a concierge that helps scope the decision in front of you.

The concierge, the reference material and the Library are free and require no account. The Institute earns its keep through private advisory engagements — a readiness assessment, drafting or reviewing a firm AI policy, designing a verification step — billed as fixed fees agreed in writing beforehand. It takes no money from any company that builds or sells legal AI: no sponsorship, referral fee, affiliate commission, revenue share or paid placement.

Because a ranking would be worth less than it appears and would cost more than it is worth. Product rankings age within months, cannot be validated against your matters, and could not be published credibly by anyone connected to people building in this space. What travels is the evaluation criteria — how to test grounding, what to ask about data handling, how to run a pilot that can actually produce a negative result — and that is what the Institute publishes instead.

Less than the marketing implies. Stanford researchers testing the leading commercial legal research products reported hallucination rates between roughly 17% and 33% depending on the product, and — the finding that matters more — examined vendor claims that retrieval-augmented architectures prevent hallucination and concluded those claims were overstated. Vendors have published improvements since. The rate is not zero for any product, which is the only figure that would let a firm skip verification.

No, and the distinction is not a formality. Rules of professional conduct are adopted state by state and they diverge. The Institute describes what ABA Formal Opinion 512 addresses and what firms do in practice; it does not tell you what your jurisdiction requires, whether particular conduct is permissible, or whether something would be sanctionable. Anything involving an actual incident belongs with counsel qualified in professional responsibility, immediately.

With what is actually happening rather than with the policy document. Prohibition without a sanctioned alternative reliably produces shadow AI — the same use, moved somewhere you cannot see it — so the first useful step is establishing what tools are already in play and for what. From there the order that tends to hold is: one approved tool people can actually use, a verification step for anything leaving the firm, then the written policy. A policy nobody can recall under deadline pressure is not governance.