7 Questions to Ask a Law Firm About AI Before You Hire

AI is now standard equipment at most law firms. LexisNexis, Westlaw, Harvey, and a growing field of specialized tools have moved from pilot programs to core infrastructure — and firms are marketing that fact in every pitch deck and intake form. For an individual or small-business owner interviewing attorneys, this creates a real problem: nearly every firm will say it uses AI. Few will volunteer what that means for how your matter is handled, who reviews the output, or what happens to your confidential information.

The questions to ask a law firm about AI during a consultation are not technical. You do not need to understand large language models or vector databases. You need to ask specifically enough to distinguish a firm that has genuinely integrated these tools from one that licensed a subscription and moved on. The following seven questions do that work — and the answers reveal more about a firm’s actual practices than most clients expect.

Why the Questions to Ask a Law Firm About AI Have Changed in 2026

A year ago, a client asking about AI tools might have been satisfied with a general answer about “leveraging technology to serve clients better.” That answer is no longer sufficient — because the ethics rules have caught up with the technology.

ABA Formal Opinion 512, issued July 2024, established the first national ethics framework governing attorney AI use, covering competence, confidentiality, client disclosure, supervision of AI output, and billing. As of early 2026, forty-seven state bars had issued formal AI ethics guidance of their own, most anchored to those same obligations. And a new catalog of AI policies from 128 law schools, assembled in August 2026, reflects how rapidly formal standards are hardening across the profession.

This means attorneys now have specific, enforceable obligations around how they use these tools. Clients have a corresponding right to ask about them — and a firm that treats these questions as intrusive is telling you something important.

1. Which Specific AI Tools Do You Use, and for What Tasks?

This is the foundation question. A firm that has genuinely integrated AI into its practice can answer it without hesitating — naming platforms (LexisNexis Protégé, Westlaw Precision, Harvey, Casetext, or others), describing which tasks they apply to (legal research, contract review, document drafting, intake processing), and distinguishing where the tool is used from where it is not.

A vague answer — “we use AI to stay current” or “we have access to the latest technology” — tells you the firm has acquired something without a clear plan for how it works. The legal AI market has matured into a modular landscape: LexisNexis’s Protégé platform, for instance, lets firms assemble specific AI capabilities by practice area rather than deploying a single undifferentiated tool. A firm that can describe its configuration has made deliberate choices. One that cannot has not.

2. Who Reviews AI Output Before It Reaches Me?

This question goes directly to professional responsibility. ABA Formal Opinion 512 is explicit: under Model Rules 5.1 and 5.3, attorneys must supervise AI output exactly as they supervise work by a junior associate. AI-generated research, analysis, or draft language that goes to a client unreviewed is an ethics violation — not a workflow choice.

The practical stakes became clear in 2023, when a federal judge sanctioned two New York attorneys for submitting a brief containing AI-generated citations to cases that did not exist. A Texas attorney faced the same consequence in 2024. These are not isolated incidents of bad faith — they are examples of what happens when supervisory protocols are absent.

The answer you want: a named attorney or specific role is responsible for reviewing AI output on every matter, with a defined workflow. “The AI assists, but I sign off on everything” is acceptable. “The tool is quite accurate” is not.

What to listen for:

• Strong answer: A named attorney or a dedicated review role is responsible for all AI-generated output before it reaches the client; the workflow is defined and consistent across matters, not applied case-by-case.
• Weak answer: “The tool is quite accurate” or “we review anything important” — these describe intent, not a process. Supervisory obligations under Model Rule 5.3 require a protocol, not an attitude.

3. Does Your AI Use Affect My Bill — and How?

Most clients never ask this question. They should.

ABA Formal Opinion 512 is clear on fees: they must remain reasonable and consistent with actual time spent, even when AI compresses that time. A research task that formerly took four hours and now takes one should cost what one hour costs — the efficiency gain cannot simply remain with the firm.

In practice, the picture is uneven. A 2026 Bloomberg Law survey found that AI has done little to reduce law firm billable hours in aggregate — firms tend to increase capacity rather than reduce per-matter cost. Firms that have restructured their pricing around efficiency gains are the exception, and worth finding. For a closer look at how those gains land — and where they remain inside the firm — see our AI legal services guide for small businesses in 2026.

Ask directly: Does AI use factor into your time estimates on this type of matter? Do you offer fixed-fee arrangements for work where AI handles a significant portion of the research or drafting? A firm that has thought carefully about this will have a concrete answer. A firm that has not will say something like “the savings get passed along” without being able to explain how.

What to listen for:

• Strong answer: The firm can explain how AI affects time estimates for this type of matter and has a concrete mechanism — revised estimates, fixed-fee options, or a stated billing policy — for reflecting efficiency gains in what you pay.
• Weak answer: “The savings get passed along” without a description of how, or a claim that AI does not affect billing at all. Under ABA Formal Opinion 512, that claim requires an explanation, not an assertion.

4. What Is Your Firm’s AI Policy — and Is It in Writing?

A written AI policy is a meaningful differentiator. It indicates the firm has moved from ad-hoc tool use to institutional practice — that partners have agreed on which tools are appropriate, how output is reviewed, and how AI use is disclosed to clients.

The ABA’s checklist for responsible AI use in law firms, published in 2026, and growing bar-association guidance reflect how quickly expectations are hardening. A firm that has done this internal work will produce its policy without prompting. If a firm has no written policy, ask how they handle situations where different attorneys on your matter use AI tools differently. The absence of a clear answer is itself a data point.

5. What Happens to My Confidential Information Inside Your AI Platforms?

Client confidentiality extends to the tools used in representation — this is settled under Model Rule 1.6. Uploading a client’s financial records, contracts, or trade secrets to an AI platform that retains input data for model training or fine-tuning is a genuine ethics exposure, not a theoretical one.

ABA Formal Opinion 512 recommends that attorneys obtain informed client consent before processing confidential information through AI tools, and specifies that boilerplate engagement letter language is not sufficient. The questions worth asking: Does your AI platform retain client input data? Is there a zero-data-retention agreement with the underlying AI vendor? What security certifications — SOC 2 Type II, ISO 27001 — does the platform hold?

A firm that can answer these specifically — naming the vendor agreements and data handling terms — has done the due diligence the ethics rules require. One that cannot has not.

What to listen for:

• Strong answer: The firm names the AI vendor, references a data-processing or zero-retention agreement with that vendor, and can identify relevant security certifications — SOC 2 Type II or ISO 27001 — by name.
• Weak answer: “We use a reputable platform” or general compliance claims without specifics. Meeting one regulatory standard (HIPAA, for instance) does not address whether client input data is retained for model training.

6. Has Your Firm Encountered AI Errors, and How Were They Handled?

This question is intentionally direct. A firm that has used AI tools seriously over any meaningful period has almost certainly encountered at least one instance where AI output was wrong, incomplete, or misleading. A firm that claims otherwise is either very new to AI or has not been paying close attention.

What matters is not the error itself — every tool fails at some point — but the protocol. How was the error caught? By whom? What workflow change followed? A firm that has processed an AI error transparently and built a stronger review protocol around it is more trustworthy, not less. That transparency is a reliable proxy for how seriously the firm takes professional responsibility obligations.

7. Will You Tell Me When AI Generates a Significant Portion of My Work Product?

This is the disclosure question, and it is the clearest test of a firm’s client orientation.

ABA Formal Opinion 512 establishes that attorneys should inform clients when AI use materially affects the cost, timing, or methodology of representation. The underlying logic is straightforward: if a task that formerly took six hours now takes one because of AI, and that change affects your invoice or the character of the deliverable, you are entitled to know. As of early 2026, forty-seven state bars had issued ethics guidance anchored to similar disclosure principles.

Some firms disclose proactively — in the engagement letter or at the outset of a consultation. Others wait to be asked. A firm that tells you upfront what AI tools will be used on your matter, and how they affect deliverables and billing, is telling you something important about how it treats the client relationship. A firm that frames the question as overreach is telling you something too.

What to listen for:

• Strong answer: Disclosure is proactive — written into the engagement letter or raised at intake — covering which tools will be used on your matter and how AI use affects deliverables, timing, and billing.
• Weak answer: Disclosure only happens if the client specifically asks, the firm treats the question as unusual, or the answer is “we follow all applicable rules” without specifying what those rules require in practice.

What the Answers Reveal

A firm that answers these seven questions specifically, without defensiveness, and without retreating to marketing language has done the institutional work that genuine AI adoption requires: written policies, trained attorneys, vendor agreements that protect client data, and pricing that honestly reflects efficiency gains.

A firm that gives vague or evasive answers has not done that work — yet. That is not automatically disqualifying; AI integration is genuinely difficult and the profession is moving at uneven speeds. But it is information you should have before you sign an engagement letter and before you share anything that cannot be undisclosed.

For a detailed look at how AI is reshaping legal costs for smaller clients — and where efficiency gains actually land versus where they stay inside the firm — see our AI legal services guide for small businesses in 2026. For a broader overview of when and how small businesses should engage outside counsel, our guide to small business legal services covers the foundational questions before any technology vetting begins.