The short answer
Yes, lawyers can use AI tools at intake — the ethics rules already assume you delegate. Receptionists, answering services, paralegals, and offshore intake teams have taken first calls for decades. An AI receptionist sits in the same category: a nonlawyer assistant performing a nonlegal task under your supervision.
In July 2024 the ABA issued Formal Opinion 512, its first opinion squarely on generative AI. It doesn't bless or ban any tool — it maps the existing duties you already owe (competence, confidentiality, communication, candor, fees) onto AI use. The same reasoning applies when the AI answers your phone.
The duties, applied to phone intake
Competence — Rule 1.1
You must understand enough about the tool to use it responsibly. For an AI receptionist that means knowing what it asks callers, what it says about your firm, and where its limits are — the same way you'd review a human receptionist's script. You can't set it and forget it.
Confidentiality — Rule 1.6
Callers share sensitive facts on a first call. Two practical questions follow: does the vendor protect recordings and transcripts appropriately, and does it use your calls to train models? A vendor that trains on your client data or has weak security is a real problem, whichever side of the AI debate you're on.
Communication — Rule 1.4
Formal Opinion 512 says lawyers should consider whether to tell clients that AI is being used. Whether disclosure is required depends on context — but as a practical matter, an intake agent that identifies itself as automated handles this cleanly, and most callers don't care who answers as long as the help is real.
Supervision — Rules 5.1 and 5.3
This is the operative rule for any delegation: you're responsible for the assistant's conduct as if it were your own. Practically: the agent must never give legal advice, must know when to escalate, and you should review what it's telling callers — read the transcripts periodically, like you'd spot-check a new hire.
No unauthorized practice — Rule 5.5
The agent's job is intake: who is calling, what kind of matter it is, the details your intake needs for that practice area, and how to reach them. What intake may ask varies — a personal injury template can ask about the accident itself, while a criminal template deliberately never asks what happened. In every case it should never evaluate a claim, predict an outcome, or answer legal questions. A well-built legal intake agent is explicitly instructed to decline those, same as a human receptionist would be.
The practical checklist
- Ask the vendor directly: are my calls and transcripts used to train models? Get it in writing (a DPA is the usual vehicle).
- Review the intake script the way you'd review a hire's training — the questions asked, the claims made about your firm, the escalation triggers.
- Confirm the agent can't dispense legal advice and declines to discuss case merits.
- Check your state's recording-consent rules before recording calls — see our state-by-state consent guide.
- Spot-check transcripts and summaries periodically, especially in the first weeks.
- If your state bar has issued AI-specific guidance, read it — a handful of states have weighed in beyond Opinion 512.
A note on honesty
This guide is general information, not legal advice, and it was written by the firstcase team — we build an AI receptionist, so read it as an interested party. Opinion 512 and your state's rules are the primary sources; if intake AI is a close call in your jurisdiction, that's a question for ethics counsel, not a vendor's website.