When the Client's First Call Comes From an Agent

A prospective client types one sentence: "Find me three family lawyers near me who can help with a property settlement, compare what they charge, and ask each one if they can see me this month." Then she goes to lunch.

When she comes back, the work is done, or most of it. The agent has searched, read the firms' websites, looked at reviews, drafted an enquiry to each of three firms and is waiting for her to say "send". She picks from the replies.

Nothing in that scene requires technology that doesn't exist. Whether people actually use it this way for legal matters, and how often, is something nobody can yet tell you. I can't either. But the direction is clear enough to think about now, while the cost of preparing is low.

From answers to actions

For the past two years the conversation about AI and legal discovery has been about answers: whether ChatGPT, Gemini or Perplexity mentions your firm when someone asks a question. I've argued that even that framing is often wrong. Ranking in ChatGPT is the wrong mental model, and a mention is not the same as a recommendation.

Agents add a further step. The discovery funnel most firms are built around looks like this: search → shortlist → client contacts firms. With an agent, it compresses to: client briefs the agent → agent shortlists, compares and contacts firms → client picks.

The client still decides. But the shortlist, the comparison and the first contact can be done by software. The first "person" at your firm who reads the enquiry may be answering a machine.

What agents can verifiably do today

It's easy to overstate this, so here is what I could verify from the providers' own documentation as at 30 September 2026. I've kept to capabilities rather than a platform tour.

They can research and act inside a browser. OpenAI introduced ChatGPT agent in July 2025. It researches across public websites and can take actions such as filling in forms, and OpenAI says it is designed to ask permission before actions with real-world consequences, such as a purchase or sending an email. Anthropic's Claude in Chrome, available on paid plans since December 2025, takes actions in the browser with site-level permissions and confirmation for sensitive actions. Perplexity describes the assistant in its free Comet browser as able to take actions on a user's behalf, and Manus lists a browser operator and email handling among its tools.

At least one can phone a business. In July 2025 Google announced that Search could call local businesses on a user's behalf to check pricing and availability, starting with services like pet groomers, rolling out to all Search users in the US. Google has since added "agentic calling" to check whether nearby stores have a product in stock, also US-only. According to TechCrunch's report of the launch, every call begins by announcing that it's an automated system calling from Google on behalf of a user. Businesses can manage whether they take part through their Business Profile settings.

What I could not verify. I found no primary source saying any of these products is designed for, or routinely used for, contacting law firms. Google's calling feature is documented for categories like pet grooming and retail, not legal services. I found no reliable data on how many people use agents to find professional services. Some products' claims (including reported new agent modes elsewhere) I couldn't confirm from a primary source, so I've left them out.

So the honest summary is: the building blocks exist and are consumer-grade. The legal use case is plausible, not demonstrated.

How an agent might choose a firm

This section is inference, not evidence. No one outside the providers knows how these systems weigh firms, and the providers haven't said.

But the mechanics suggest some things. An agent working through a browser can only compare what it can read. So I'd expect it to favour firms whose websites make the basics easy to extract:

• Scope. Does the firm clearly say it does this kind of matter, in this place?
• Price signals. Fixed fees, a fixed-fee first meeting, or at least a clear statement of how costs work. If the client asked the agent to "compare what they charge", a firm that says nothing gives the agent nothing to compare.
• Availability. Can the agent find out whether the firm can help this month without a phone call?
• Third-party signals. Reviews, directory listings and how consistently the firm is described across sources.
• A route to contact that the agent can actually use.

None of that is new advice. What changes is who is doing the reading. A person tolerates a vague practice page and picks up the phone. An agent may simply move to the next firm.

Intake is the new battleground

My view is that if agent-mediated enquiries become common, they will be won or lost less in marketing than in intake. Five things matter.

Response speed. An agent comparing three firms will report back whatever it has when the client returns. A firm that replies in two hours is in the comparison. One that replies in two days may not be.

Published scope and price signals. A clear page saying what you do, for whom, and how you charge is readable by a person and a machine alike.

Forms a bot can complete. Long forms, mandatory phone fields, dropdowns that don't match real matters, and CAPTCHAs will stop some agents. I'm not suggesting you remove spam protection. I am suggesting you look at your form through the eyes of someone who wants to describe a matter in a paragraph and ask a question.

Answering without a phone call. Many firms convert by phone, and for good reason. But if the only route is "call us", an email-first agent can't proceed. Offer a written reply path that can at least confirm scope, next steps and how fees work.

How you respond when the enquiry says it comes from an AI assistant. This will happen. Google's calls already announce themselves. My suggestion is a short, standard written reply: confirm whether the firm handles this type of matter; explain how initial fees work; state that the firm will need to speak with the person directly before advising or agreeing to act; and give the next step. No legal advice to the agent. Treat it as a real enquiry from a real person who has chosen a new way to reach you.

The professional questions no one has answered yet

I'm flagging questions here, not giving answers. The rules differ between jurisdictions, and I'd want regulators and professional bodies to speak before anyone treats this as settled.

• Who is the client? The person who briefed the agent, obviously. But when does an agent's enquiry create any duty to them, and how do you verify who you're dealing with?
• Confidentiality. Matter details passed through an agent have been shared with a third-party provider before they reach you. What does that mean for how you respond, and what you ask for?
• Advertising and solicitation. If a firm replies to an agent with a tailored pitch, is that advertising, or a response to a request? If firms start deploying their own agents to respond, where does solicitation begin?
• Volume and spam. If agents make contacting five firms effortless, intake could see more enquiries, more duplicates and more automated noise. Can your team triage that?
• Conflicts. An agent may contact several firms with the other side's name in the message. Receiving that information can create conflict issues before anyone has spoken to the client.

What stays human

Agents may do the searching and the first contact. I doubt they'll do the choosing, at least for matters that matter.

People pick a lawyer on trust: whether this person understands their situation, will tell them the truth and will be there when it goes wrong. That judgement usually forms in the first conversation. An agent can get your firm into the shortlist. It can't hold the client's nerve in a first consultation. The firms that do well will be the ones that make it easy to reach that conversation and then make the conversation good. In 20 years of working with law firms, that part hasn't changed.

What to do now

Keep it cheap and sensible:

1. Read your own website as a machine would. Can someone establish, from text alone, what you do, where, for whom, and how you charge?
2. Publish price signals where you can. Fixed fees, a fixed-fee first meeting, or a plain explanation of how costs work.
3. Test your enquiry form. Can it be completed with a paragraph and an email address? Does it work on the first try?
4. Measure response time. From form submission to first reply, by practice area.
5. Write the standard reply for an enquiry that says it comes from an AI assistant, and agree it with whoever looks after risk.

The AI visibility checklist covers the wider audit.

What to avoid

Don't rebuild your site for bots. The things agents appear to need (clarity, scope, price, a working contact route) are the things people need. If a change helps only a machine, be sceptical.

Don't buy "agent optimisation". There is no public evidence about how agents select law firms, so no one can credibly sell you a method. Ask what it's based on.

Don't advise the agent. Answer the enquiry, not the legal question, until you've spoken to the person.

Don't assume it's here at scale. I've found no data showing how much legal work is being sought this way. Prepare cheaply; don't reorganise.

Confidence summary

Could your firm even notice?

The simplest test of readiness is whether you'd know an agent had contacted you at all. Most firms wouldn't. At Practice Proof, which I founded, we look at intake readiness as part of the broader picture of AEO for law firms: whether a firm's website, forms and first response work for the person, or the software, on the other end.

The interesting question isn't whether AI will mention your firm. It's whether, when the agent arrives, your firm wins the comparison.

FirmRanker was founded by Dan Toombs. Practice Proof, also founded by Dan, provides digital and AI visibility services to law firms.