AI Client Intake for Small Law Firms: Answer Every Lead Without Creating a Prospective Client by Accident

Updated 2026-10-21 · Some links may be affiliate links. Disclosure

Most small firms lose new clients before a lawyer ever hears about them. Clio’s Legal Trends Report found that 79% of prospective clients expect a reply within 24 hours, yet 67% of firms didn’t respond to email inquiries and 48% didn’t respond to phone inquiries, as reported by LawSites (LawSites, Aug 20, 2026). Many inquiries also arrive after hours, when nobody is there to answer.

AI intake tools now answer the phone, the website chat and the inbox around the clock, ask your screening questions, check names against your conflicts list, and book a consultation when a lead looks like a fit. That’s useful, but an AI intake agent also talks to people who may become prospective clients under Rule 1.18, so it has to be set up the way you’d train a new intake assistant: it says what it is, it asks only what you need, and it never gives legal advice.

This guide covers intake, from first contact to booked consultation. For the work after you’re hired, see our contract drafting guide and our legal research and citation-checking guide.

Not legal advice. The ethics rules below are summarized from the ABA Model Rules and from state bar guidance; your state’s version of each rule controls, so check it before you go live.

Ranked picks for AI client intake

Why this order: The fastest win is answering every inquiry, and all four do that. Clio Grow comes first because Grow AI covers chat, email and phone in one system, flags conflicts for your review, and charges per hired matter rather than per conversation (Clio help). The catch is that you need Clio Grow, whose add-on price isn’t public. Smith.ai is the cheapest way to start: a free plan with 25 real calls a month, published per-call pricing, no contract, and integrations with Clio (Smith.ai), Lawmatics and Filevine (Smith.ai). It’s phone-first, so it doesn’t replace a CRM. Lawmatics is the deepest intake CRM, but you’ll pay for at least three users plus AI usage, and one lead can count as more than one AI Outcome. Intaker publishes no pricing at all. A $80/month starter figure still circulates on review sites, but it comes from an Intaker blog post dated January 4, 2023 (Intaker blog), so don’t budget on it.

Step 1: Measure where leads leak today

For four weeks, log every new inquiry: channel (phone, web form, chat, email, referral), time received, time of first human response, whether a consultation was booked, and whether the person hired you. Then test-shop your own firm: call at 7 p.m. on a weekday, submit your web form on a Saturday, and send an email to your general inbox. Most firms find one leaky channel. Point AI at that one first.

Step 2: Decide what the AI may and may not do

Write it down before you configure anything. The guidance is consistent across the sources:

1. It must say it’s AI. The Florida Bar says AI chatbots that talk to clients or third parties must include a disclaimer that the chatbot is an AI program, not a lawyer or law firm employee (Florida Bar Ethics Opinion 24-1).
2. It collects facts and never gives legal advice. Opinion 24-1 applies the rules for nonlawyer intake staff to chatbots: identify the nonlawyer status, limit questions to gathering facts, give no legal advice about the matter or the fee agreement, and send legal questions back to a lawyer. It also says a lawyer may not delegate to AI anything that could be the practice of law (Florida Bar).
3. You supervise it. Your supervisory duty under Model Rule 5.3 extends to AI tools such as chatbots, and you should review every script the chatbot might use and keep it from saying anything that could form an attorney-client relationship (ABA Labor & Employment Law magazine).
4. Screen out people who already have a lawyer. Opinion 24-1 suggests screening questions that limit the chatbot’s conversation if the person is already represented (Florida Bar).

Check the defaults. Lawmatics says Merlin Engage will never claim to be human, but it confirms it’s AI when a lead asks (Lawmatics help). Florida’s opinion calls for telling people up front, so put the disclosure in the agent’s first message. Smith.ai says businesses can customize the greeting to introduce the AI transparently (Smith.ai).

Step 3: Design intake questions around Rule 1.18

Under Model Rule 1.18, someone who consults with you about possibly hiring you is a prospective client. Even if you never take the case, you can’t use or reveal what they told you, and if they tell you something “significantly harmful,” you may be blocked from representing the other side.

1. A chatbot can create a consultation. The ABA comment says a consultation is likely if you specifically invite people to submit information about a potential representation without clear and reasonably understandable warnings and cautionary statements that limit your obligations, and they respond (ABA Model Rule 1.18, Comment 2). The State Bar of Arizona’s guidance for firms with chatbots: make clear it’s a chatbot, not a lawyer; don’t invite confidential information; and include a disclaimer (State Bar of Arizona).
2. Ask only what you need to decide. Comment 4 says that to avoid acquiring disqualifying information, limit the initial consultation to what reasonably appears necessary to decide whether to take the matter (ABA). In practice: contact details, practice area, location, how they heard of you, and the names of every other party, then stop. Leave the story for the consultation, after conflicts clear.
3. Put the disclaimer where people type. A short line above the chat box and in the first message, for example: “I’m an automated assistant for Smith & Lee, not a lawyer. I can’t give legal advice. Please share only your contact details, the type of matter and the names of other people or businesses involved. Don’t send details of your situation yet. Contacting us doesn’t make us your lawyers.” Have a lawyer approve the final wording for your state.

Step 4: Vet the vendor for confidentiality

ABA Formal Opinion 512 applies Rule 1.6 confidentiality to generative AI and notes that Rule 1.18(b) extends similar protection to prospective clients’ information. Before information goes into an AI tool, you must evaluate the risk that it’s disclosed to or accessed by others, inside or outside the firm (ABA Formal Opinion 512). Opinion 24-1 adds three questions from the cloud-computing opinions (Florida Bar):

• Does the vendor have an enforceable obligation to keep the data confidential and secure, and to notify you of a breach or a subpoena?
• What are its security measures and its limits on liability?
• Does it keep your data after you cancel, or claim rights to it?

What the vendors say: Clio says Grow AI only accesses information from your account, encrypts all data and never uses it to train external models (Clio). Smith.ai records and transcribes every call and says recordings and transcripts are stored encrypted (Smith.ai). Ask Lawmatics and Intaker the same three questions in writing. Since phone agents record calls and recording-consent laws vary by state (some require every party’s consent), have the greeting announce the recording.

Step 5: Turn on one channel and script it

Start with the channel Step 1 showed is leaking.

• Phones: Smith.ai lets you train and test the AI on simulated calls in its Quality Studio before going live, and its free plan includes test calls (Smith.ai). Clio’s voice agent answers calls when your team is unavailable, gathers case details and can book a consultation during the call. It’s US-only for now (Clio help). Clio’s offer page lists unlimited phone-agent minutes as an introductory offer through October 30, 2026 (Clio), so ask what minutes cost after that.
• Web chat and email: Grow AI answers common questions in website chat, monitors your connected inbox, replies to gather missing details, and routes every lead to your Clio Grow inbox with a summary and a fit assessment (Clio help). Intaker handles chat and texting, and auto-replies to Google LSA message leads (Intaker).

Give the AI your practice areas, the counties or states you serve, your consultation fee and your hours. Also give it a list of things it must hand to a person: anyone who mentions a court date or deadline in the next two weeks, an arrest, or a safety concern. Those are your firm’s calls to make, not the AI’s.

Step 6: Screen for fit and run conflicts before anyone books

1. Fit scoring is a suggestion. Clio says Grow AI scores leads against your past cases and checks for conflicts, but doesn’t decide which clients you take; your team reviews and decides (Clio). Lawmatics’ Merlin Qualify returns recommendations such as “Chase Hard,” “Refer Out” or “Reject” (Lawmatics help). A lawyer or trained staff member should confirm every decline.
2. Know how each conflict check works. Clio’s Grow AI flags potential conflicts for your review (Clio help). Smith.ai’s Conflict Checker searches your Clio, Lawmatics or Google Sheets contacts (current clients, former clients and leads) for a person’s name or a business name, not both in one search, and it’s an extra-cost add-on to Smith.ai’s live receptionist plans (Smith.ai docs). Lawmatics lists basic conflict checking on Essential and advanced conflict checking on Premium (Lawmatics).
3. Run conflicts before the consultation, not after. Book only when conflicts are clear. Matching names is a start; a person still reviews related entities, maiden names and spelling variants.

Step 7: Let AI follow up, within limits

Most of the value is in chasing leads who go quiet. Grow AI follows up on incomplete intake forms, unsigned agreements and unbooked consultations (Clio help). Lawmatics’ Engage agents reach leads by text, email or both, with one as the fallback (Lawmatics help). Three limits:

• Stop when they say stop. Model Rule 7.3(c) bars solicitation when the person has made known a desire not to be solicited, or when it involves coercion, duress or harassment (ABA Model Rule 7.3). Cap the number of follow-ups and end the sequence on any “no thanks.”
• Honor text opt-outs. Under the FCC’s rule, a reply such as “stop,” “quit,” “end,” “revoke,” “opt out,” “cancel” or “unsubscribe” revokes consent to automated texts, and you may send only one confirmation afterwards (47 CFR 64.1200(a)(10), (a)(12)).
• AI voices count as artificial voices. The FCC ruled in 2024 that AI-generated voices are “artificial” under the TCPA, so outbound AI calls need prior express consent unless an exemption applies (FCC 24-17). Keep AI calling to inbound calls and to callbacks people asked for.

Step 8: Hand off, and close the loop with declined leads

Every AI conversation should end in your system with a summary a person reads the same business day. Smith.ai sends new leads to your Clio Grow inbox after each call (Smith.ai), and Grow AI leads land in your lead inbox with a summary and fit assessment (Clio help). If you decline, have a person send a short, plain letter saying you won’t represent them, without commenting on the merits of the case, so nobody believes you’re their lawyer.

Step 9: Measure after 30 days

Compare your Step 1 log with the first 30 days: time to first response, share of after-hours inquiries answered, consultations booked, hires, and cost per hire (Grow AI’s per-hire fee, Smith.ai’s per-call charges, or Lawmatics’ Outcomes). Read 20 transcripts, too. Look for any advice-like answers, missing AI disclosures, or callers who told their whole story before conflicts ran. Fix the script before you add a second channel.

Vendor question list

• How does the agent identify itself as AI in its first message, and can we edit that wording?
• Can we limit the questions it asks, and stop it from collecting case details before conflicts clear?
• How is the conflict check run, against which records, and who sees the result?
• Do you train models on our data, how long do you keep recordings and transcripts, and what happens to them when we cancel?
• Will you notify us of a breach or a subpoena for our data?
• What’s the full price: platform, users, per-call or per-outcome fees, phone minutes, add-ons and cancellation terms?

Quick checklist

• Four-week inquiry log and a test-shop of your own firm
• Written list of what the AI may and may not do; legal questions go to a lawyer
• AI disclosure and Rule 1.18 disclaimer in the first message and above the chat box
• Intake questions limited to contact details, matter type, location and other parties’ names
• Vendor confidentiality terms reviewed (training, retention, breach notice); recordings announced
• Conflicts run and reviewed by a person before any consultation is booked
• Follow-ups capped; “stop” and “no thanks” end the sequence; no cold outbound AI calls
• Every lead summary read by a person the same business day; declines sent in writing
• Results and 20 transcripts reviewed after 30 days

Compare and calculate

• AI time-savings ROI calculator: Hours saved, dollar value and payback for an AI tool, with wage defaults for your profession.

Tools in this guide

• Clio Grow (with Grow AI): Small firms on Clio that want chat, email and phone intake with conflict flags in one system.
• Intaker: High-volume firms buying Google Local Services Ads or social leads that need instant chat and text replies.
• Lawmatics (with Merlin AI): Growing, marketing-heavy firms that want a full intake CRM with AI qualification and follow-up.
• Smith.ai AI Receptionist: Firms that mainly lose leads on the phone and want to start free on any practice management system.

Related playbooks

• AI Legal Research for Small Law Firms: How to Use It Without Citing Fake Cases
• Best AI for Contract Drafting at Small Law Firms (and How to Use It Ethically)
• AI Receptionists for HVAC, Plumbing & Electrical: Stop Losing Jobs to Missed Calls

More AI tools and playbooks for Small & Solo Law Firms →