Can AI Legally Handle Client Intake? Model Rule 5.5, explained

Last reviewed 31 July 2026  ·  Rule text quoted from the American Bar Association pages linked below, checked on the date shown.

Can a law firm legally use AI for client intake?

Yes, within a boundary that is short to state. Software may gather facts, acknowledge an inquiry and schedule a human. It may not give legal advice, because a firm that lets a nonlawyer system do that is assisting the unauthorized practice of law. The rule that governs it is Model Rule 5.5, and the rule that governs the supervision is 5.3.

Nothing in the Model Rules bans automation. There is no rule that says a first response must be typed by a person, and no rule that says a form acknowledgement has to wait for office hours. What the rules constrain is content and supervision, not the tool.

What does ABA Model Rule 5.5 actually say?

Rule 5.5 prohibits a lawyer from assisting the unauthorized practice of law. Paragraph (a) reads: “A lawyer shall not practice law in a jurisdiction in violation of the regulation of the legal profession in that jurisdiction, or assist another in doing so.” For anyone buying intake software, the words that matter are assist another.

Read literally, the rule is not about robots. It is about a lawyer enabling someone unlicensed to do lawyer work.¹ A system that answers a stranger’s question about whether they have a claim is doing lawyer work, and the firm that pointed it at the public is the one doing the assisting.

One caveat that matters more than it looks. The ABA Model Rules are a template. Every state adopts its own version, some renumber, and some add requirements the model does not contain. Read your state’s rule, not this page.

Where exactly is the line between intake and legal advice?

The line is whether the message contains a conclusion. Collecting facts is intake. Applying law to those facts, in any direction, is advice. A message that says “when did this happen?” is safe. A message that says “then you still have time” is not, and the difference between them is one sentence in a script.

The last row surprises people. Declining feels like the safe direction, because nobody is being promised anything. It is not safe: a person told by software that their matter is not worth pursuing may act on that, and the firm said it.

The message-by-message version of this, including what the first text may and may not contain, is in the intake script guide.

Who supervises the software under Model Rule 5.3?

A named person at the firm. Rule 5.3 requires partners and lawyers with managerial authority to make reasonable efforts to ensure the firm has measures giving reasonable assurance that a nonlawyer’s conduct is compatible with the lawyer’s professional obligations, and imposes the same duty on anyone with direct supervisory authority over that nonlawyer.

Paragraph (c) sets out when the lawyer is personally responsible for the conduct: if they ordered or ratified it with knowledge, or if they had managerial or supervisory authority, knew of it at a time when the consequences could still be avoided, and did not take reasonable remedial action.²

That is the practical test, and it is easy to run. Ask whoever owns the tool to paste the current script into an email. If they have to ask the vendor, supervision is happening somewhere other than at your firm.

What did the ABA say about AI specifically?

It issued Formal Opinion 512, “Generative Artificial Intelligence Tools”, on 29 July 2024, the first formal ethics guidance on lawyers’ use of these tools. It works through competence under Rule 1.1, confidentiality under Rule 1.6, communication under Rule 1.4 and fees under Rule 1.5, and says lawyers must fully consider their ethical obligations before using them.

The opinion is notably firm on self-learning tools, warning that their design “by its very nature raise[s] the risk that information relating to one client’s representation may be disclosed improperly, even if the tool is used exclusively by lawyers at the same firm”, and that informed client consent is needed before entering representation information into them.³⁴⁵

For intake specifically that is a procurement question, not a script question, and we covered it separately in the confidentiality guide. Note also Rule 1.18: a person who consults a lawyer about the possibility of forming a client-lawyer relationship is a prospective client, and information learned from them may not be used or revealed even if no representation follows.⁶

What does compliant intake automation look like in practice?

Boring, short and fully written down. The system says four things and stops: the message arrived, who it arrived at, one or two questions, and how to reach a human or stop the messages. Everything past that waits for the callback. A firm that holds that line does not need to argue about Rule 5.5 at all.

• A fixed, written script that a lawyer at the firm has approved and can produce on request.
• No conclusions of any kind — not about liability, value, deadlines, or whether the firm will take the matter.
• A named supervisor who owns the script under Rule 5.3, not a shared inbox.
• An always-available human route and an always-available stop instruction.
• State advertising labels applied where the firm’s jurisdiction requires them.
• A one-click off switch the firm controls itself.

Firms sometimes ask whether a script that limited is worth automating. That is the wrong comparison. The alternative is not a better conversation at two in the morning. It is silence until Monday.