Nothing in this article is legal advice, and that is rule one for a law firm's AI agent too: information about your firm and its process, never advice on anyone's matter. Have your own counsel review your disclaimers and intake flow for your jurisdiction.
A law firm's website gets its most valuable visitors at the worst possible times: someone searching for a family lawyer at 10pm, a business owner reading about contract disputes on a Sunday, an injured person comparing firms after a long shift. If the only ways to reach you are a phone number and a contact form, most of those people move on to the next firm.
An AI chatbot for law firms closes that gap. Trained on your own website and documents, it answers practice-area questions, runs structured intake, triages the matter, books the consultation, and hands everything to your staff, without ever giving legal advice. Here is how to set one up, what it must never do, and how to measure it.
Why law firms lose leads after hours
Legal problems do not keep office hours. People research lawyers when the problem is in front of them: after the accident, after the letter arrives, after the argument. That moment usually falls in the evening or on a weekend, when your reception desk is empty.
Three things make it harder for firms than for most businesses:
- Speed decides who gets the consultation. Prospects rarely contact one firm. Whoever responds first with something useful usually gets the first meeting.
- Contact forms feel like a void. Describing a sensitive situation into a blank text box, with no idea who reads it, is a big ask, and many visitors abandon it.
- Phone-only intake filters out the hesitant. Many potential clients want to ask a few careful questions in writing before they commit to a call.
An AI agent answers at the moment of intent, whatever the hour. We cover the pattern in how after-hours AI support turns questions into loyalty.
What an intake AI agent actually does
Scoped correctly, the agent's job is intake and information, not law. Five tasks:
- Answering practice and process questions. "Do you handle landlord disputes?" "How do your fee arrangements work?" The agent answers from your own pages and documents, so it matches what your site says.
- Deflecting repeat FAQs. Office hours, locations, languages, what to bring to a first meeting. These questions interrupt your paralegals today, and they do not need to.
- Running structured intake. Name, contact details, location, matter type, short description, urgency, plus the opposing party's name so your staff can run a conflict check before booking time.
- Triaging by matter and urgency. Routing rules send each inquiry to the right practice group, and anything mentioning a court date or deadline is flagged for immediate human follow-up.
- Booking the consultation. The agent points qualified visitors to your booking page, and through Zapier it connects to 9,000+ apps, so captured inquiries flow into your calendar or CRM automatically.
A static form gives you a name, a paragraph, and a wait. We compare the two in AI agent versus contact form for lead capture.
What a law firm chatbot must never do
This is where legal differs from every other vertical. Treat these as hard rules:
- No legal advice, ever. "Do I have a case?" and "what should I do next?" depend on facts, jurisdiction, and judgment. The agent says plainly that it cannot advise, then offers a consultation with an attorney.
- No implied attorney-client relationship. The agent states, in its answers and in a visible disclaimer, that chatting does not create an attorney-client relationship and that nothing it says is legal advice.
- No fishing for sensitive detail. A website chat is not privileged. Intake needs contact details and a short description, not a confession, so the agent asks for the minimum and discourages sensitive narratives before a human is involved.
- No deadline math. Limitation periods and filing deadlines are where a wrong automated answer does real harm. Anything deadline-shaped is routed to a person.
- No unreviewed claims about results. Bar advertising rules differ by jurisdiction, so have your own counsel review the agent's source content as they would your website copy.
With Dante, scoping is mostly a content exercise: the agent answers only from content you give it, plus a standing instruction to decline advice questions. For more, see how to train an AI chatbot to handle compliance.
Setting it up on your own content
- Start free. Dante's free plan is $0 with no credit card: one AI agent, up to 700 onboarding credits, then 100 credits a month, where one credit equals one message.
- Train it on what you already publish. Point it at your website, then upload your fee guide, what-to-bring checklist, and consultation process. A working agent takes about 60 seconds; budget an hour to review its answers.
- Write the boundary into the content. Add a short document stating what the agent must decline (advice, deadlines, case assessments) and the handover language to use. That puts your compliance position in writing.
- Test it like an anxious prospect. Ask about fees and logistics, then ask "do I have a case?" and confirm it declines and offers an attorney.
- Embed it with one script tag. A single snippet works on HTML, Webflow, WordPress, and Squarespace. The steps are in how to add an AI agent to your website.
- Keep it current. Paid plans start at $40 per month in USD, and knowledge re-crawls run weekly on Advanced and daily on Pro. Full details are on the pricing page.
Handing conversations to your staff
An intake agent is only half a system; the other half is what your team does with what it captures. Dante's built-in human handover passes the full conversation to your staff, so a prospect never repeats their story. Make the workflow explicit: deadline-flagged conversations reach a person the same day, new-matter inquiries land in your CRM with matter type attached, and someone owns the morning review of overnight transcripts, including who calls each prospect back.
That review habit is also your compliance backstop: no conversation becomes a client matter without attorney eyes on it first.
Measuring intake conversion
Treat the agent like any intake channel and give it a funnel: conversations started, contact details captured, consultations booked, matters opened. Tracked weekly, those numbers tell you whether it earns its place.
- Compare against your form baseline. Run the agent alongside your existing contact form and compare captured inquiries per week. Keep whichever combination wins.
- Watch the after-hours share. Inquiries captured outside office hours are the clearest measure of leads you were previously losing outright.
- Read transcripts weekly. Questions the agent could not answer become content fixes, and questions it should have declined become boundary fixes.
Frequently asked questions
Is an AI chatbot for law firms confidential?
A website chat is not privileged, and no attorney-client relationship is formed by using it, so the agent should say both plainly. Handle transcripts under your privacy policy, collect only the intake basics, and discourage visitors from sharing sensitive detail before your team is involved.
Can an AI chatbot give legal advice?
No. A well-scoped AI agent shares general information about your firm, practice areas, fee structures, and process. When a question asks it to apply the law to someone's situation, it declines and offers a consultation with an attorney.
How do clients react to a law firm chatbot?
They react well when it is scoped honestly: instant answers at any hour, a clear statement that they are talking to an AI agent rather than a lawyer, and an easy path to a human. Frustration comes from agents that pretend to be people or dodge handover.
How much does an AI chatbot for law firms cost?
Dante's free plan is $0 with no credit card: one AI agent, up to 700 onboarding credits, then 100 credits a month. Paid plans are $40, $120, and $400 per month in USD with 2 months free on yearly billing. One credit equals one message.
Dante is one good option here, not the only one, and the criteria above apply to any tool you evaluate: it should train on your own content, decline advice questions reliably, hand over with full context, and report its funnel. To see how it handles your firm's questions, start free and test it before any client ever sees it.