Law Firm Intake With AI: Faster First Contact, No Accidental Legal Advice

It's 10:40 on a Thursday night. Someone fills in the contact form on a small firm's website: "My landlord changed the locks while I was at work. My things are inside. What can I do?" The firm reads it on Friday at 9:15, replies at 11, and gets no answer. By then the person has called three other firms, and one of them picked up.

There's a second way this goes wrong, and it's newer. The firm has a chatbot on its website. It replies within seconds, sympathetically, and adds: "Based on what you've described, you likely have a strong case for illegal eviction." Nobody at the firm has seen the facts. The chatbot has just given legal advice in the firm's name.

Good intake avoids both failures: it answers fast, and it never says anything a lawyer wouldn't sign. This post is organised as the questions I'd expect a managing partner to ask before trying it.

What does intake actually involve?

Strip away the software and intake is five steps: someone gets in touch, the firm checks for conflicts, someone works out whether the matter fits (practice area, jurisdiction, urgency, whether the firm can help), a consultation is booked, and an engagement letter follows if the firm takes it on.

Only one of those steps needs legal judgment: deciding whether the firm can and should help. The rest is gathering information, checking it against records, and scheduling. That's where the time goes, and it's also where most firms lose people: not because the lawyer said no, but because nobody got back to them in time.

Where can AI help without crossing the line?

Everywhere information is being collected, sorted or scheduled:

• Answering immediately, at any hour, with a clear statement of what happens next and when a person will be in touch.
• Asking structured questions that match the matter type: a landlord dispute needs different facts than a car accident or a will.
• Identifying the practice area and jurisdiction, so the inquiry reaches the right person, or gets a polite referral if the firm doesn't handle it.
• Spotting urgency signals: a court date, a deadline mentioned in a letter, someone in danger, someone in custody.
• Collecting names for the conflict check before collecting detailed facts (more on why below).
• Booking the consultation and sending a list of what to bring.
• Writing a summary for the lawyer, so the consultation starts with the facts already on the table.

Why collect names before facts?

This is the part firms most often get backwards. Under ABA Model Rule 1.18, information a prospective client shares is protected even if the firm never takes them on. And if the firm receives information that could be significantly harmful to that person, it can be disqualified from representing the other side in the same matter.

So an intake assistant that cheerfully collects every detail of a divorce before anyone has run a conflict check can create a real problem. If the other spouse is already a client, the firm now holds information it shouldn't have.

The fix is simple and belongs in the design: collect names of the parties and the basic matter type first, run the conflict check, and only then gather the detailed facts. An assistant can hold that order more consistently than a busy receptionist can. It can say: "Before you tell me more, I need the names of everyone involved so we can make sure we're able to help you. This is standard for every law firm."

How does the flow look?

Which questions should it ask?

The questions depend on the matter type, and writing them down is the most useful hour a firm can spend on intake, with or without AI. A starting point:

The assistant asks, records and flags. The urgency column doesn't trigger advice. It triggers a person.

What does the lawyer receive?

A one-page summary, not an assessment. It might look like this:

Matter: Residential tenancy, possible lockout. Jurisdiction: state and county given. Parties: tenant (inquirer), landlord (individual, name given), property manager (company, name given). Conflict check: clear, checked Friday 9:05. Key dates: lease signed March 2025; rent paid through end of month; locks changed Thursday. Urgency: tenant has no access to belongings including medication; flagged for same-day call. Documents: lease (uploaded), photo of the notice on the door (uploaded). What they want: access to the apartment and belongings. Not yet known: whether any court order exists.

Notice what's absent: no opinion on whether the lockout was lawful, no likely outcome, no deadline calculation. Those are the lawyer's call. The summary saves the first ten minutes of every consultation and makes sure the lawyer walks in knowing the urgent part.

What should the prospective client see?

The hardest moment in automated intake is when someone asks, directly, "Do I have a case?" The answer has to be honest, warm, and firm:

I can't answer that, because I'm an automated assistant and not a lawyer. Only an attorney can tell you after looking at the details. What I can do is make sure an attorney sees everything you've told me today, and book you a consultation. Because you mentioned your medication is inside, I've marked this as urgent and someone will call you this morning.

It's the same thing a good receptionist says. The assistant should say it every time, word for word if needed, and never slide into "it sounds like" or "usually in cases like this".

Is this compatible with the ethics rules?

Yes, if you design it as assistance to your staff rather than a substitute for your lawyers. The ABA's Formal Opinion 512 on generative AI tools, issued in July 2024, walks through the duties that apply: competence, confidentiality, communication with clients, supervision and reasonable fees. For intake, three stand out.

Confidentiality. Prospective client information is covered. Before any of it goes into an AI tool, you need to know where it's processed, whether the provider stores it, and whether it's used for training. Choose providers with a contract that rules out training on your data.

Supervision. An intake assistant is a nonlawyer assistant in everything but form. Someone at the firm is responsible for what it says, which means reviewing its instructions, reading a sample of conversations every week, and fixing anything that drifts toward advice.

Honesty about what it is. The DoNotPay case is a reminder that claiming an AI can do a lawyer's job is a consumer protection problem, not only an ethics one. The assistant should say what it is in its first message.

Several state bars have issued their own guidance on AI, and it varies. Check yours before launch.

Which tools do firms use for this?

Most small firms already have part of the stack. Intake and CRM tools such as Clio Grow, Lawmatics and the intake features in MyCase, PracticePanther or Smokeball handle forms, pipelines and scheduling. Virtual receptionist services such as Ruby and Smith.ai answer the phone, and several now add AI to the mix. The AI layer I'd add sits between them: it runs the structured conversation, enforces the names-first order, flags urgency, and writes the summary into your practice management system.

Start with one practice area where inquiries are frequent and fairly standard, such as landlord and tenant, personal injury, estate planning or employment. Measure response time to first contact and how many inquiries turn into booked consultations, before and after.

When isn't it worth it?

If your firm works mostly on referrals from other lawyers, and new matters arrive by a phone call from someone you know, automated intake adds little. If your practice is complex commercial work where every inquiry needs a partner's attention from the first minute, the same applies. And if you get fewer than a handful of new inquiries a week, a staff member with a good intake checklist and a firm rule to reply within two business hours does the job well.

A rule for every first contact

Answer every inquiry fast, check conflicts before collecting the story, and let nobody but a lawyer say what the story means. An assistant can do the first two better than a busy front desk. The third stays with you.

If new inquiries are slipping through at your firm, tell me which practice management software you use and which matter types bring in most inquiries. I'll suggest how a names-first intake could fit in. For a related intake problem in a different profession, see how accounting firms can stop chasing client documents, and how insurance brokers prepare renewal calls without losing the personal touch.