Most Am Law firms are writing their 2027 innovation strategy right now. The memo usually reaches the executive committee before the holidays, and it decides what the firm can attempt next year. I've read several drafts this fall, and most of them are a list of tools: a chat assistant, a drafting add-in for Word, a research product, a pilot or two, and the renewal date for each.
A list of tools tells the partnership what the firm will pay for. It doesn't say what the firm expects to be able to do in December 2027 that it can't do today, and that's what partners will judge the program on a year from now.
The money is already moving. Law firms grew technology spending 9.7% and knowledge management spending 10.5% in 2025, according to the 2026 Report on the State of the US Legal Market from Thomson Reuters and Georgetown Law (summary at LawSites). Here's what I think the strategy behind that spending should record.
Start with the questions the firm wants answered
Write down five or six questions a partner should be able to ask by the end of 2027 and get an answer that cites the firm's own documents. Make them specific to the firm's practices. What position have we taken on this indemnity cap across every deal in the last five years, and why? Which of this client's agreements carry a change-of-control clause that this transaction triggers? Who in the firm has handled a matter like this one, and what did they decide?
Questions like these do two jobs. They tell the technology team what the firm's data has to support, and they give the executive committee a test it can apply in twelve months. If partners can ask them in December 2027 and get cited answers, the program worked.
Decide what the firm will own
Every AI contract a firm signs decides where its work product goes: the documents it processes, the corrections its lawyers make, and the structure it builds around its own matters. A strategy should say in one paragraph what the firm intends to own when each contract ends. That covers the firm's data, the model of its own work, and the right to export both in a form another system can use. It should also say what the firm won't accept, such as a vendor training on firm data, even when the data is anonymized.
This is the question I hear most often from general counsel and CIOs once the vendor has left the room. It's much easier to settle in a strategy memo than in the middle of a renewal. How to evaluate a legal AI vendor covers the contract questions in more detail.
Fix client and matter attribution first
Every useful AI answer in a law firm depends on knowing which client and which matter a document belongs to. Attribution is what makes ethical walls enforceable, what lets an answer be scoped to one client, and what lets the firm learn from its own history. In most firms a meaningful share of documents can't be tied to a matter with confidence, because they arrived by email, were saved to a personal folder, or came from the client without a matter number.
The source of truth usually exists already, in the document management system, the matter management system or the time and billing system. The strategy should name it and commit the firm to tagging every record against it before anything is loaded into an AI system. I explained why this rarely requires a data lake in Why most law firms don't need a data lake before AI.
Name who steers it
At many firms, AI decisions still go through a technology committee whose job is choosing and supporting tools. That committee does necessary work, and it isn't set up to own strategy. The memo should name a small group that does: one or two practice leaders who will use the results, the CIO, the head of knowledge or innovation, and the general counsel. That group meets monthly, reviews progress against the questions in the first section, and decides what to stop as well as what to start. How IT and the business can build a law firm's agentic strategy together describes how that partnership works day to day.
Plan for lawyers who build their own tools
Some of your lawyers are already building small applications with AI: a script that checks a closing set against the signing checklist, an agent that drafts a first-pass chronology from a production. That will grow in 2027 whether or not the strategy mentions it. The strategy should give IT a way to say yes safely, with a governed environment where business-side builders work against approved data, with logging, access control and a review step before anything touches client work.
Price the program so success doesn't raise the bill
Seat licenses and metered usage both grow with adoption, so a program that succeeds costs more every quarter. That's a hard conversation with a partnership that's also watching AI compress billable hours. The strategy should state the firm's preference for platform spend, fixed pricing wherever it can be negotiated, and it should model what happens to cost if usage doubles.
Put privilege and client guidelines in the plan
In February, Judge Rakoff of the Southern District of New York held in United States v. Heppner that a defendant's exchanges with a consumer AI tool were protected by neither attorney-client privilege nor the work product doctrine. Clients are asking their outside counsel how AI is used on their matters, and the answers need to be consistent across the firm. The strategy should say which tools are approved for client work, how the firm records AI use on a matter, and who answers a client's questions about it.
Show the partnership a result within two quarters
AI programs rarely lose support because the technology fails. They lose it when partners stop seeing a return. Pick one practice group and one of the questions from the first section, and commit to answering it for that group by mid-2027. A visible result in one practice earns the patience the rest of the program needs.
How we approach it at Reframe
We start every engagement with the first and third items on this list: the questions the firm wants answered, and the source of truth for every client and matter. From there we build a model of the firm's own work inside the firm's own cloud. Every answer cites the document it came from, and every change our agents propose is approved by a person before it's recorded. The firm owns the result.
Seven questions to test your draft
- What will a partner be able to ask in December 2027 that they can't ask today?
- What will the firm own when each AI contract ends, and can we export it?
- Which system is the source of truth for client and matter, and what share of our documents is tagged to it?
- Who owns this strategy, and how often do they meet?
- How does a lawyer who wants to build something get a safe yes?
- What happens to our costs if usage doubles?
- Which practice group sees a result first, and by when?
If your draft can't answer one of these yet, that's the section to write next.
Drafted with Reframe Editorial Agent, an AI employee at Reframe. 