Someone rings your firm at 6pm about a dispute with their landlord. They get voicemail, leave a rushed message, and by 9am they've called two other firms. One of them answered. Law firm intake is where a lot of good matters are quietly lost, and it's also one of the clearest places for an AI agent to help, as long as a lawyer stays in charge of every decision that matters.
This post covers what good intake looks like, how to set up a matter pipeline in HubSpot so an agent can work it, what the agent does at each stage, and the sign-off points that always stay with a person. It applies to conveyancers too, and I'll cover where their work differs.
Good intake isn't about speed alone. It's about capturing the right facts once, so nobody has to ask again. For most firms that means:
An agent can collect all of this through website chat, a form or a follow-up email, at any hour, and in a consistent format. That alone is a big improvement on a voicemail and a sticky note.
We build agents on HubSpot because it gives them a structured map to follow. Each matter becomes a record (a custom object or HubSpot's Services object, depending on how your firm works), associated to the client contact, the other parties, any companies involved and the responsible lawyer. If you want the basics of how that hangs together, our plain-English guide to pipelines, objects and associations walks through it.
A typical intake pipeline looks like this:
Because each stage has a clear meaning, the agent always knows where a matter sits and what should happen next.
The agent responds within minutes, thanks the person, explains the next step and asks for any missing details. It creates the contact and matter records and links them. It says plainly that it can't give legal advice and that a lawyer will review the enquiry.
The agent searches HubSpot for every party named and lists any matches on the record for a person to review. It doesn't clear the conflict. Clearing a conflict is a professional judgement, and a lawyer makes it.
Once cleared, the agent offers times from the right lawyer's calendar, sends a confirmation with what to bring, and reminds the client the day before.
The lawyer decides whether to take the matter and approves the costs disclosure. The agent sends the approved documents, tracks whether they've been signed and chases politely if they haven't. That chasing works exactly like the document-chasing agent we use across professional services.
Every firm we work with gets a written list of hand-off rules before anything goes live. For legal intake, these always stay with a person:
The agent never says "you have a strong case" or "you should settle". If a prospective client asks for advice, the agent explains that a lawyer will need to answer that, books them in or flags the matter, and passes on what was asked. This is the core difference between a scripted chatbot and an agent, which I covered in chatbot or AI agent? An agent can take real actions, so the rules about which actions it may take matter even more.
We also start every agent in draft mode. A team member approves each message before it's sent. An agent only earns more freedom once 95% of its work has been approved unchanged for two weeks, and the sign-off points above never move, however good it gets.
Conveyancing intake is more predictable than litigation, which makes it a great fit. The agent can capture property and party details, request the contract and identity documents through a secure upload, track key dates on the matter record (cooling-off, finance and settlement), and keep buyers and sellers informed about which stage their file is at. Reviewing the contract, advising on special conditions and anything to do with settlement funds stays with the conveyancer or lawyer.
We roll intake agents out using the same four steps every time. Context: check the data, pick one job (usually after-hours enquiries), agree what done means and measure today's cost, such as how many enquiries go unanswered for more than a business day. Action: HubSpot's own agents first, custom only where they stop, tested on real past enquiries. Coordination: people approve, the agent acts. Result: a monthly scorecard showing what the agent did, what it saved, what it cost and the net return.
If you're wondering how much of your team's week goes on this kind of admin, where your 10 hours of admin actually go is a useful read before you start.
Ready to stop losing good matters to voicemail? Book an AI Readiness Audit and we'll map your intake and pipeline, or go to Build my solution if you already know what you need. You can also see the full range of agents we build.