AI for Small Law Firms: Where It Saves Real Time (and Where It Doesn't)

Small law firms are being sold AI from every direction right now. Every legal tech vendor has added an AI feature, every conference has an AI panel, and every newsletter promises that firms who don't adopt will be left behind. Meanwhile, the solo practitioner or five-lawyer firm is still trying to figure out a much simpler question: where does this actually help me, and where will it get me in trouble?

I come at this from two directions. I practiced law for twenty years, including running my own firm, so I know what a small practice's week actually looks like. And since 2025 I've been training AI models on legal reasoning and drafting as an independent contractor for several companies building these tools. That combination gives me a clear view of both what these systems can do and where they reliably fall short. Here's my honest assessment.

Where AI genuinely saves time

First drafts of routine documents. Engagement letters, demand letters, routine correspondence, discovery requests built from a standard set, basic motions with well-established formats. AI is very good at producing a competent first draft from a clear instruction. The lawyer still has to review and own every word, but turning a blank page into an 80-percent draft in minutes is a real gain.

Summarizing large volumes of material. Deposition transcripts, document productions, medical records, long email chains. AI can produce useful summaries and chronologies that help a lawyer get oriented quickly. This is where many small firms see the biggest time savings, because document-heavy matters are exactly where a small team gets buried.

Client intake and communication. Structuring intake questionnaires, organizing the information a client provides, drafting plain-English explanations of a process for a client who's anxious and confused. I've built intake automation for post-judgment collection work that handles the repetitive front end of a matter so the lawyer can focus on judgment calls.

Brainstorming and issue-spotting. Asking an AI to list the arguments the other side is likely to make, or the issues a fact pattern might raise, is a useful way to pressure-test your own thinking. It won't replace your judgment, but it can catch something you missed.

Where AI does not save time, and can cost you

Legal research you don't verify. This is the one that has already produced sanctions. General-purpose AI tools can generate case citations that look perfect and don't exist, or real cases cited for propositions they don't support. Every citation must be checked against a primary source. If the verification takes as long as the research would have, you haven't saved anything. You've just added a risk.

Anything requiring current, jurisdiction-specific law without a reliable source behind it. Rules change, local practice varies, and AI tools trained on general data don't reliably know the current version of your local rules or the judge's standing order. Use legal-specific tools that cite sources you can check, and check them.

Strategic judgment. Whether to settle, how hard to push, how a particular judge or opposing counsel is likely to react, what the client really needs as opposed to what they say they want. These are the parts of the job that clients actually pay for, and they're the parts AI handles worst.

The guardrails that matter

Confidentiality comes first. Client information should never go into a consumer AI tool that may retain or train on your inputs. Use tools with enterprise-grade data protection, read the terms, and set a firm-wide rule about what can and cannot be entered. Your ethical duty of confidentiality doesn't pause because a tool is convenient.

Supervision is the second guardrail. AI output is work product prepared by a non-lawyer, and it needs the same review you'd give a first-year associate's draft. That means reading it critically rather than skimming it. A useful habit: have someone check the AI's work who didn't write the prompt.

Third, be transparent where it matters. Courts increasingly have rules about disclosing AI use in filings, and clients may have expectations about it in billing. Know the rules that apply to you and follow them.

How to start without betting the practice

Pick one repetitive, low-risk task that eats real hours every week. Intake, routine correspondence and summarizing records are good candidates. Choose a tool with appropriate data protections, run it alongside your existing process for a few weeks, and measure whether it actually saves time after you account for review. If it does, standardize it and move on to the next task. If it doesn't, you've lost very little.

That incremental approach is far more valuable than a firm-wide "AI transformation." Small firms win with AI the same way they win everything else: by being practical, careful and focused on what actually moves the work forward.

Considering AI for your firm?

Book a consultation and I'll help you find the tasks where it saves real time, with guardrails that protect your clients and your license.