How to Use an AI Intake Agent Without Crossing the UPL Line
A plain-English guide for attorneys on keeping an AI intake agent on the right side of unauthorized-practice-of-law rules: what it may do, what it must never do, and how to vet a vendor.
By The Embido Team
An AI intake agent can answer your firm's website day and night, qualify prospects, and book consultations while you sleep. It can also, if built or configured carelessly, wander into giving legal advice, and that is where a convenience turns into a liability. This guide explains where the line sits and how to stay on the right side of it.
What "unauthorized practice of law" means for a chatbot
Every jurisdiction reserves the practice of law to licensed attorneys. The exact definition varies, but the core idea is consistent: applying legal knowledge to a specific person's situation, advising them on their rights, or predicting how their matter will turn out is practicing law. A piece of software has no license, so when it does those things on your firm's behalf, the exposure lands on the firm.
The good news is that the most valuable part of intake, capturing the right information and getting the prospect onto your calendar, is not the practice of law at all. It is what a skilled intake coordinator does every day.
The bright line: gathering information vs. giving advice
Keep one distinction in mind and most of the risk disappears.
On the safe side of the line, the agent may:
- Ask what kind of matter the prospect has, when it happened, and where.
- Explain, in general terms, what your firm does and which practice areas you cover.
- Collect names, contact details, and a description of the situation.
- Offer available times and book the consultation on your calendar.
- Tell the prospect what to bring or expect at that first meeting.
On the wrong side of the line, it must never:
- Tell the prospect what they should do about their legal problem.
- Apply the law to their specific facts ("based on what you've described, you have a strong claim").
- Predict an outcome, a settlement range, or odds of success.
- Suggest they do or refrain from doing something with legal consequences.
The test is simple: gathering facts and scheduling is administrative; interpreting those facts through the law is advice. A safe agent stays firmly in the first category and deflects anything in the second back to "that's exactly what the attorney will go over with you."
Three guardrails every deployment needs
A UPL-safe intake agent is not just a friendly chatbot with a disclaimer bolted on. Look for three things working together.
-
Advice refusal by design. When a prospect asks "do I have a case?" the agent should recognize the request and redirect to a consultation rather than answer it. This has to be built into how the agent responds, not left to chance in a general-purpose model.
-
A clear, visible disclaimer. Every conversation should carry your own language stating that the agent is not a lawyer, does not provide legal advice, and that no attorney-client relationship is formed by using it. The disclaimer belongs in the chat itself, where the prospect actually reads it.
-
A conflict pre-check. Before the agent encourages someone to book, it should screen the prospect, and the adverse party, against your existing clients so you are not inadvertently inviting a conflict. This stays advisory: it surfaces a signal for the attorney, it never clears or declines a matter on its own.
Keep a human, and a record, in the loop
UPL rules exist alongside your broader duties of competence and supervision. Two habits keep you covered. First, a licensed attorney reviews every qualified lead and owns the decision to take the matter, the agent hands off, it does not decide. Second, the system keeps a log of what was asked and answered, so if a question about a conversation ever comes up, you can see exactly what the prospect was, and was not, told.
A short vendor checklist
Before you point any AI agent at your firm's website, get clear answers to these:
- How does the agent refuse or redirect questions that ask for legal advice?
- Does it ever apply law to a prospect's specific facts? (The answer should be no.)
- How and where is the "not legal advice / no attorney-client relationship" disclaimer shown?
- Does it run a conflict pre-check before encouraging a consultation?
- Is every reply logged in a way you can review later?
- Can you set the practice areas, disclaimer, and transfer rules yourself?
If a vendor cannot answer these plainly, that is your answer. Embido's Client Intake Agent is built around exactly these guardrails, advice refusal, your own disclaimer, an advisory conflict pre-check, and a full audit trail, so the safe path is the default path. For how we protect the underlying client data, see our security overview.
Frequently asked
- Can an AI intake agent commit the unauthorized practice of law?
- The tool itself does not hold a license, but how you deploy it matters. If an automated agent gives legal advice, applies law to a prospect's specific facts, or predicts an outcome, it can create UPL and malpractice exposure for the firm that runs it. Kept to gathering information and booking consultations, it stays on the right side of the line.
- Does an AI intake conversation create an attorney-client relationship?
- It should be built so it does not. A well-designed agent states plainly that it is not a lawyer, does not give legal advice, and that no attorney-client relationship is formed by the chat, and it runs a conflict pre-check before encouraging a consultation.
- What should I ask a vendor about UPL safety?
- Ask how the agent refuses advice-seeking questions, whether it applies law to facts, how the disclaimer is presented, whether it runs a conflict pre-check, and whether every reply is logged. If the vendor cannot answer these clearly, treat that as a red flag.
See Embido work on your own website
We set Embido up on your site for you: intake agent, phone lines, and training on your firm. Book a consultation and go live in days.
Book a consultation