Short answer: automate the capture and the assembly, keep the judgment. Intake is a good first use of AI in a law firm because so much of it is moving information, not making decisions. It is a bad place for full automation, because the decisions it feeds — whether to sign the client, how to frame the case, what to promise — are the ones your license exists for.
This page is the plain version of how we explain it to firm owners. It applies whether you run a two-attorney practice or a growing personal injury firm.
The four things people mean by "automating intake"
Vendors use the same phrase for very different products. Separating them matters, because they carry very different risk.
1. A form or chatbot on your website. It collects names and answers to scripted questions. Cheap and predictable. It cannot hear urgency, it cannot pull context from an existing relationship, and it stops being useful the moment a caller's situation drifts off-script.
2. An answering service. Humans take messages after hours, usually from a script. Reliable, but they don't know your matters, and everything they collect has to be re-typed into your systems by your staff the next morning.
3. An AI voice agent. Software answers the phone and converses. The pitch is 24/7 coverage. The failure mode shows up regularly in lawyer communities: it commits to things — appointments, advice, expectations about the case — that no one at the firm approved.
4. A managed intake assistant. The model we build: the assistant listens to calls, reads email and texts, assembles what it gathers into a packet a human reviews, and drafts the follow-up. A person at your firm reads it, edits it, and makes the call back. Nothing goes to a prospective client without a human pressing send.
If a vendor cannot tell you in one sentence who approves what their system sends, treat it as category three, whatever the brochure says.
Where AI genuinely helps
After-hours capture. Missed calls are the oldest intake problem in law. An assistant that answers, collects the story, checks it against existing records, and hands your morning person a complete summary changes the first hour of the day. The firm still decides who gets a call back and what is said.
Packet assembly. Intake for a personal injury matter means gathering provider records, dates, insurance details, and prior correspondence into something an attorney can actually read. An assistant that works from the firm's own documents and drafts a timeline — with citations back to the source records — removes hours of paralegal assembly without removing the paralegal's judgment.
Status and follow-up drafts. "Where does my case stand" messages deserve better than silence or a template. An assistant that drafts the update from what is actually in the file — for a staff member to approve and send — keeps clients informed without putting words in the firm's mouth.
In all three, the pattern is the same: the AI does the reading, gathering, and drafting. People decide.
Where it should not go
- Signing decisions. Whether the firm takes a case is a legal and business judgment. No intake system should make it.
- Legal information to prospects. The moment a caller asks what their case is worth or what they should do, a human takes over.
- Confidential calls without a policy. If the assistant records or transcribes, your team should know exactly where those transcripts live, who can read them, and what your obligations are — the same duties of competence and confidentiality the ABA's guidance on generative AI (Formal Opinion 512) asks firms to apply whenever AI touches client work.
- Anything that pretends to be a person. Prospects should know they are talking to an assistant, and how to reach a human immediately.
Questions to ask any intake vendor
- Who reviews what the system sends, and can it send anything without a person?
- Where does caller and client information end up — which systems, which data centers, whose models? If the answer to "can it run on hardware in our office" is a blank stare, you have your answer.
- What does it do when a caller describes something off-script or urgent?
- How does it connect to the practice management system we already run — Clio, Filevine, whatever it may be — or does it create another silo our staff has to check?
- What happens to our data if we stop?
A vendor that manages the system with you, rather than selling you a login, will have answers. A self-serve tool usually will not.
A realistic first step
Do not automate intake for the whole firm on day one. Pick one cohort — new personal injury inquiries, say — and one workflow: capture after-hours calls, assemble the packet, draft the callback summary, attorney reviews before anyone dials. Run it for thirty days. Your staff will tell you quickly whether the drafts are useful or noise, and you will know whether to widen the circle.
That is the approach we take with firms: learn how the firm works, connect carefully to the systems it already keeps its records in, and leave every decision with your attorneys. If you want to talk through whether it fits your practice, reach out here — and if it is not a fit, we will say so.
More on our overall approach: AI for law firms at Normal Technology.