How to write AI-use language into your engagement letter.

Why the engagement letter is the right place

ABA Formal Opinion 523, covered by Massachusetts Lawyers Weekly on June 8, 2026, addresses client restrictions in engagement letters. The detail to take from it for this purpose is simple: the engagement letter is the recognized place to set the terms of the relationship, including what the client may restrict and what you, the lawyer, set out up front.

Pair that with the broader 2026 trend of disclosing AI use to clients, and the conclusion writes itself. If clients are increasingly going to ask how their lawyer uses AI, the engagement letter is where you answer it cleanly, in advance, on your own terms. You set the expectation once, in writing, instead of fielding the question awkwardly after a client reads a headline.

A caution before we go further. This is about putting sensible AI-use language into your agreement. It is not a claim that any rule forces you to use AI or to disclose it in a particular form. Treat the clause as good practice for managing the client relationship, not as compliance with a specific direct mandate.

The reason to do it anyway, mandate or not, is that the engagement letter is the one document the client reads carefully and keeps. It is where they form their picture of how you work. Putting your AI-use position there, in your own measured words, means the client's first impression of the subject is one you chose, rather than one they assembled from a news story or a competitor's marketing. That control is worth a paragraph.

What clients actually want to know

When a client asks about AI, they are usually asking three things underneath the one question.

• Are you still doing my work, or is a machine doing it? They want to know a lawyer remains responsible and in control.
• Is my confidential information safe? They want to know their private matter is not being fed into something careless.
• Am I being charged fairly for it? The billing question, which lives in its own place.

A good engagement clause answers the first two directly and reassuringly, and leaves the third to your billing terms. The goal is not to bury the client in technical detail. It is to make a nervous question unnecessary by addressing it before it is asked.

There is a tone question underneath the wording, and it matters more than the exact phrasing. A clause that sounds defensive teaches the client to be suspicious. A clause that sounds matter of fact, this is how a modern firm works, and a lawyer is always responsible, teaches the client that you have thought about it and have nothing to hide. The same information lands completely differently depending on whether you write it as a confession or as a standard. Write it as a standard. You are not apologizing for using good tools. You are telling a sophisticated client how their work gets done, and doing so before they have to wonder.

A ready-to-adapt clause, with the reasoning

Here is a plain starting clause. Adapt it to your practice, your jurisdiction, and your actual tools. The reasoning behind each line follows, because you should never paste language you cannot explain.

> Use of technology and artificial intelligence. In providing legal services, this firm may use software tools, including artificial intelligence tools, to assist with tasks such as drafting, research, and document review. A lawyer at this firm remains responsible for all work performed and reviews it before it is provided to you or filed. We take reasonable steps to protect the confidentiality of your information when using these tools, consistent with our professional obligations. If you have questions or specific concerns about the use of these tools in your matter, please raise them with us so we can discuss them.

Line by line:

• "May use software tools, including artificial intelligence tools." This is honest disclosure without over-promising or naming products that may change. It tells the client AI may be involved.
• "A lawyer remains responsible for all work and reviews it." This is the heart of it. It tells the client that responsibility and judgment stay with a person, which is what they actually care about.
• "Reasonable steps to protect confidentiality, consistent with our professional obligations." This signals you take confidentiality seriously without making a guarantee you cannot keep. What is actually safe to put into a tool is its own question, covered in What is safe to put into AI.
• "Raise concerns so we can discuss them." This invites consent and conversation. Under the spirit of Opinion 523, a client may want to restrict certain uses, and this line opens that door rather than pretending the client has no say.

Handling the client who wants to restrict AI

This is where the engagement letter earns its keep. Some clients, after reading your clause, will want to limit or forbid AI use in their matter. ABA Opinion 523, as reported, deals with exactly this territory of client restrictions in the engagement relationship.

The practical posture: take the restriction seriously, decide whether you can honor it, and document the answer. If a client says do not use AI on my files, you either agree and note it, or you have an honest conversation about what that means for cost and timing. What you do not do is agree in the letter and quietly ignore it. A restriction you accepted and then breached is a far worse position than one you negotiated honestly up front. The clause that invites the conversation is what makes that negotiation possible before the work begins.

One more practical point. A restriction is only meaningful if your team can actually follow it. If a client forbids AI on their matter and you accept that, everyone who touches the file needs to know, and your tools need to be set up so the prohibition holds in practice and not just on paper. This is where the engagement clause connects back to internal supervision. A promise you made to a client is only as strong as the workflow that keeps it, which is why the language in the letter and the method in the office have to match. Promise what you can deliver, then build the office so you deliver it.

Where this ends and billing begins

Keep one boundary clear. This briefing is about what goes into the engagement agreement: the disclosure of AI use, the assurance that a lawyer stays responsible, and the opening for client consent or restriction. It is the relationship and consent layer.

How you charge for AI-assisted work, whether you bill saved time, how it appears on the invoice, is a separate question with its own ethics. We handle that in Billing for AI-assisted legal work. Set the engagement terms here. Set the billing terms there. Keeping them distinct in your own mind keeps both clean in the agreement.

The skill under the clause

A clause is only as good as the practice behind it. Telling a client a lawyer reviews all AI-assisted work means nothing if you do not actually have a reliable method for that review. The engagement language is a promise. The method is how you keep it.

That method, using AI on real legal work while staying responsible, confidential, and defensible, is the thing worth building. The Leveraged Attorney teaches it from the ground up for practicing lawyers, and the two minute course quiz will point you to the right starting place for your practice.