AI for Law Firms

How I’d look at AI in a law firm

A small or mid-sized firm runs on a handful of systems: practice management, document storage, email, a research service, billing, and the phones. Nearly everything a firm does is reading, writing, and deciding, which is exactly why AI vendors are so interested in lawyers, and why the pitches can get loud.

I start by looking at where the firm’s time actually goes. For some firms it’s intake: calls that come in while everyone is in court. For others it’s drafting routine documents, or summarizing records, or chasing time entries at the end of the month. The right tool depends on which of those is actually costing you.

Start with what you already own

• Your practice management system. Clio, MyCase, Smokeball, Filevine and others have added AI features that work on your matter data. Since they already hold your client information under agreements you’ve reviewed, they’re often the safest place to begin.
• Your research subscription. Westlaw and Lexis both offer AI research assistants built on their databases. If you already pay for one, ask your rep what’s included and what it costs to add.
• Microsoft 365 or Google Workspace. Both have AI assistants that work inside email and documents, under business terms. For summarizing and first drafts of correspondence, that may be enough.

Where AI genuinely helps

• Summarizing. Deposition transcripts, medical records, long email chains, and document productions. AI gives you a map; you still read what matters.
• First drafts. Engagement letters, routine correspondence, standard agreements, and discovery responses, drafted from your own templates and edited by an attorney.
• Intake. Answering after hours, collecting the facts you need to screen a matter, and getting a consultation on the calendar.
• Comparing documents. Spotting what changed between contract versions, or which clauses deviate from your standard.
• Research, with the right tool. Legal research products grounded in real case law can speed up the first pass. You still read the cases.

Where it’s overhyped or risky

• General chatbots for research. They can produce citations that look right and don’t exist. Courts have sanctioned attorneys for this, and some judges now require disclosure of AI use in filings.
• Confidentiality. Client information in a tool that retains it, trains on it, or shares it with third parties is an ethics problem. Read the terms, not the marketing page.
• Intake bots that sound like lawyers. An AI receptionist should gather information and schedule. It should not evaluate a case or suggest what someone should do.
• “Replace your associates.” AI can make a first draft faster. It cannot exercise judgment, and supervising its output is still the attorney’s job.
• Billing. Charging hourly for time a tool saved, or billing the client for your AI subscription without an agreement, can create problems. Decide your approach on purpose.

Questions to ask any vendor

1. Is client data used to train your models, and how long is it retained? Can we get that in the contract?
2. Does it integrate with our practice management and document systems, or does someone upload files by hand?
3. For research: what database is it grounded in, and does it link to the actual sources?
4. Can we control which staff can use which features and see what was done?
5. What is included in our current plan, and what is a paid add-on?
6. How do we export our data if we leave?

Where I’d start

For most firms: a written AI policy that reflects your state bar’s guidance, a business-grade assistant for summarizing and drafting, and a look at what your practice management and research tools already include. If intake is where you lose matters, that’s usually the next project. Bigger commitments can wait until you’ve seen how your team actually uses the first ones.