AI Receptionist for Law Firms and Solicitors: Intake Without Advice

What does intake look like when nobody on the phone can give advice?

An enquirer rings with a problem and wants to know what to do. The agent's job is to make sure the right lawyer learns the facts quickly, and to say so honestly. A typical call runs like this:

1. The agent says it is an automated assistant, that it cannot give legal advice, and whether the call is recorded.
2. It asks which kind of matter the call is about and takes the caller to the matching branch of the script.
3. It records name, number, safe times to call, and whether leaving a message on that number is acceptable.
4. It asks for the names of other people or organizations involved, so the conflicts team can search before anyone replies.
5. It asks whether a deadline, hearing or notice is involved, and flags the call as urgent if so.
6. It explains what happens next, using only the promise you approved, such as a callback within a stated working window.

Everything else is declined politely. A caller who asks whether they have a good case hears that a lawyer needs to look at the details. That is also the correct answer for the firm, so it is not merely a limitation.

Hours change the ending, not the script. During office hours an urgent matter can transfer to a duty lawyer or the relevant team. Outside them the agent promises only the follow-up window you have approved, records the call, and sends the written summary to whoever opens the inbox first. Practice areas you do not handle get a short, courteous decline, so your team is not working through enquiries it was never going to accept.

Why do conflicts and confidentiality shape the handoff?

Names collected on a first call can matter as much as the facts. The agent therefore never answers "do you act for X?" or "are you already working against my employer?". Even confirming that a client relationship exists can disclose something confidential. It records the question and passes it to the conflicts team.

It also does not run a conflicts search itself. Matters are accepted only by a person after the check. The agent's output is a tidy record: the caller, the parties named, the practice area and the urgency flag, delivered to the people who decide.

The rules behind this are general, and you should confirm how they apply to you:

• England and Wales. The SRA Code of Conduct for Solicitors says at paragraph 6.2 that you do not act where you have a conflict of interest, and at paragraph 6.3 that you keep the affairs of current and former clients confidential unless disclosure is required or permitted by law or the client consents. The paragraphs we checked mention current and former clients, so ask your compliance officer how they apply to an enquirer.
• United States. The ABA Model Rules cover client confidentiality (Rule 1.6) and duties to prospective clients (Rule 1.18). ABA Formal Opinion 512 (July 2024) gives guidance on generative AI and confidentiality, informed consent and supervision. States adopt their own rules, so check yours.

Can it update Clio, LEAP or other practice-management systems?

Only if the vendor gives access, and that access is not the same across products. We checked the following in November 2026.

• Clio: Clio publishes developer documentation and uses OAuth 2.0, with separate portals for Clio Manage and the Clio Platform. Its Platform currently exposes the Clio Grow APIs, which is the part of its range that concerns intake. Read the Clio Manage documentation to see which records an outside app can create.
• LEAP: runs a developer console and marketplace, but the technical reference sits behind developer registration. Whether we can write a new enquiry into your LEAP account depends on that access, so we would raise it with LEAP first.
• Others: if your system has an API or webhook, a workflow in n8n or Make can usually send a record to it. We have not verified other products by name and would check yours before quoting.

If no connection is possible, the fallback is an encrypted email or secure task sent to your intake team, entered by hand. That keeps a person in the loop at the cost of speed. In a mixed setup, Clio Grow could receive new-enquiry records while the matter stays in the main system until a lawyer accepts it, though we would confirm the access first. Website Prism's founder has delivered 800+ projects over 8+ years, mostly on Fiverr, and Website Prism Ltd was incorporated in 2026. We have no documented legal-sector case studies, so expect a supervised pilot on real calls.

What would we refuse to build?

• An agent that gives legal advice, assesses prospects of success or tells a caller what to do about a deadline.
• One that accepts instructions, sends engagement letters or takes money, which are acts for a lawyer.
• One that tells any caller whether the firm acts for or against a named person.
• One that quotes fees beyond words your partners approve.
• One that handles calls involving personal safety, such as domestic abuse or a person in custody, without an immediate human handoff.
• One that hides that it is automated.

The data side deserves care too. Transcripts of legal calls are confidential material. Choose where they are stored, keep the access list short, set a retention period and decide whether to record audio at all. Callers may be speaking from a shared home or a workplace, so the script asks before leaving messages. Our post on AI and privacy covers the wider picture, and the AI receptionist service page describes the general build. This is general information, not legal advice. Please confirm with your own regulatory adviser, then contact us.