How Law Firms Are Using AI: Notes From a Legal Conference
Law firms are using AI for case research, first drafts of contracts and large document reviews, and a lawyer checks every citation and clause before it reaches a client or a court. That’s what I saw at a legal conference, and the surveys mostly back it up.
I walked in ready for some serious eye-rolls about AI.
The legal industry has a reputation: conservative, change-averse, “we’ve always done it this way.” I expected skepticism, maybe even hostility toward AI tools.
Boy, was I wrong.
How are law firms using AI?
Every booth was talking about AI. Every panel mentioned it. The hallway conversations were dominated by which AI tools firms were adopting.
The interest went well past cautious. It felt like a gold rush.
Firms were asking:
- “How do we implement this before our competitors?”
- “What are the ethical guidelines for AI-assisted research?”
- “How do we train associates on AI tools?”
Nobody seemed worried that AI would take their jobs. The worry was falling behind if they didn’t adopt.
The surveys back up the mood, with one caveat about what gets counted. Clio’s 2024 Legal Trends Report, released in October 2024, says AI usage “jumped to 79% of legal professionals, up from 19% in 2023.” The ABA’s 2024 Legal Technology Survey, run from October to December 2024, asked a narrower question and got a smaller number: 30.2% of attorneys said their offices were using AI-based tools. A lawyer trying ChatGPT on a Tuesday counts in the first number. A firm that has rolled out a tool counts in the second.
Show the data
| Firm size | Attorneys whose office uses AI tools |
|---|---|
| 500+ lawyers | 47.8% |
| All respondents | 30.2% |
| 10 to 49 lawyers | 29.5% |
| 2 to 9 lawyers | 24.1% |
| Solo practitioners | 17.7% |
Why AI is risky for law firms: hallucinated citations
Law has traits that make AI both useful and risky:
Why AI works well:
- Massive document review is time-consuming and error-prone
- Legal research involves searching large corpora for relevant precedents
- Contract analysis follows patterns that AI can learn
- Many tasks are repetitive but require reading comprehension
Why AI is risky:
- Mistakes can have serious consequences (wrong advice, missed clauses)
- Hallucinated citations are a real problem (in Mata v. Avianca, a federal judge fined two lawyers and their firm $5,000 for filing cases ChatGPT made up)
- Client confidentiality constraints on data
- Regulated profession with ethical obligations
The Middle Ground They’re Finding
What impressed me was how carefully many firms are approaching it:
AI for research, humans for advice AI does the initial case research, associates verify citations, partners review analysis. Nobody’s letting ChatGPT write the final brief.
AI for first drafts, not final drafts Contract templates generated by AI, then reviewed and customized by attorneys. Faster than starting from scratch, safer than trusting AI output directly.
Specialized tools over general AI Firms are choosing legal-specific AI tools (trained on legal corpora, designed for legal workflows) over general-purpose ChatGPT. The hallucination problem is too serious for generic tools. In the ABA survey, 74.7% of attorneys named accuracy as a major concern, more than any other risk.
The survey data is messier than the booths, though. Among AI research tools firms had adopted or were seriously considering, ChatGPT was still cited most (52.1%), ahead of Thomson Reuters CoCounsel (26.0%) and Lexis+ AI (24.3%). The specialized tools are winning the conversation faster than they’re winning the installs.
Extensive verification workflows Every AI-generated citation gets verified. Every AI-drafted clause gets reviewed. The workflow bakes in human checkpoints.
What Other Industries Can Learn
Any high-stakes industry can copy what these firms are doing. Assume AI output needs checking and build the check into the workflow. Pick a domain-specific tool where one exists. Train people on the new skills these tools need. And tell clients and regulators when AI is involved.
The Speed Surprised Me
I expected legal to be 5 years behind tech in AI adoption. It feels more like 6-12 months.
Part of this is competitive pressure. When one firm starts delivering results faster, others have to match or lose clients. Thomson Reuters’ 2025 Future of Professionals report, from a survey of 2,275 professionals across legal, tax, trade, accounting and risk and compliance, puts a number on the prize: respondents predict AI will save them 5 hours a week within the next year, up from 4 hours predicted in 2024.
Part of it is the obvious fit. Legal work involves exactly the kind of language tasks AI excels at.
And lawyers have searched research databases like Westlaw and LexisNexis for decades. An AI research tool is a new version of something they already trust.
Do law firms need an AI policy?
Yes. The questions the firms I met were still arguing about are the policy, written down before someone pastes a client file into a chatbot:
- Confidentiality: How do you use AI tools without exposing client data?
- Liability: Who’s responsible when AI-assisted work has errors?
- Ethics: What are the obligations around disclosing AI use?
- Skill development: Will associates learn to think legally if AI does the research?
Each of these argues for adopting carefully, which is what many of the firms I met were doing. Our own version of that policy is a one-page list of what an AI tool may read and what it may change, which I wrote up for connecting Claude to QuickBooks. Swap the ledger for a case file and the questions are the same.
My Takeaway
The legal industry’s AI adoption convinced me that the “AI-resistant” industries narrative is overblown.
When the benefits are clear and the competitive pressure is real, adoption happens fast, even in conservative industries.
If the profession with a reputation for “we’ve always done it this way” moved this fast, yours will too.
Law firms have the hardest version of the problem a company brain solves: a mountain of documents and strict rules about who can see which one. I wrote about the day confidential transcripts slipped past our own gate, and about what I build for executives who want answers from their firm’s files, with the source. For a lawyer, the source is the selling point. You can try it on a made-up agency’s files: ask it a question and open the document behind every line.
