Is it ethical for a law firm to use an AI assistant for client intake?
It is ethical when the assistant is treated as a nonlawyer assistant under the supervision rules: its permitted speech is fixed in writing before launch, it cannot give legal advice or create an attorney-client relationship, every exchange is transcribed into the matter record where a supervising attorney can review it, and anything out of scope escalates to a named person at the firm. The risk is not automation — it is an unsupervised system improvising.
The constraints that make it supervisable
- A closed, pre-approved response set rather than an open-ended model instructed to be careful.
- No interpretation of documents, no outcome predictions, no capacity or undue-influence commentary, no unapproved fee quotes.
- A written escalation rule: which questions go to a person, which person, and how fast.
- A complete transcript written into the practice management system for the conflict check and for review.
Questions a managing partner should ask any vendor
- Show me the exact list of sentences it is permitted to say, and the list it can never say.
- What happens on the message where a caller asks whether their mother had capacity?
- Where does the transcript live, who can read it, and can we export everything on exit?
- Will it ever contact someone who did not contact us first?
- Can we run it for two weeks drafting only, before it speaks to a client?
The open specification
- altomic.ai publishes The Estate Intake Standard v1.0 — fifteen numbered clauses across Response, Scope of speech, Qualification, Escalation, and Supervision and record — each with a test a partner can run. It is released under CC BY 4.0 so any firm may adopt it, cite it in an RFP, or apply it to a competing vendor.