Your phone is off in a courtroom. Theirs isn't.
People call a lawyer on the worst day of their month. A termination letter, a car accident, a relative in a holding cell. They search, they call the first firm that looks right, and if a solo attorney is at a hearing or in a client meeting, they get voicemail. Then they call the second firm on the list.
A missed intake call at a firm isn't like one at a plumber's, though. The reply can't just answer the question, because the question is usually legal. So the job here is narrow: give the caller a way to book a consultation, send anything substantive to you rather than answer it, and say nothing that sounds like advice or an engagement. Below is where that line sits.
The text says
- Who you are, and that the text is automated
- That it isn't legal advice and doesn't make you their lawyer
- A request not to send details of the matter by text
- Your consultation booking link, and your consult fee if you set one
- Nothing further once a reply needs a lawyer; that goes to you
It never says
- Whether they have a case, or what it's worth
- Anything about deadlines, charges, or what to do next legally
- "We can help" or "we'll take this" before a conflict check
- Questions about what happened, who's involved, or the other side
- Fee estimates for the matter itself
The right-hand list is written into the firm's intake as things the agent must never say. A second model reviews every draft against that intake before it sends. One that breaks a rule gets one rewrite; a second failure never reaches the caller and lands with you.
Three texts from a made-up firm.
Wrenfield Law doesn't exist. The shape does: every first text names the firm, says it's automated, and ends with the opt-out line, in 300 characters or fewer. The counts below are computed from the text itself.
A new inquiry, the first text
Every disclosure the firm asked for, and still under the cap. The price of all that disclosure is that there's no room for warmth. Don't try to add it.
A reply that sounds urgent
The agent doesn't try to triage a criminal matter or hand out a phone number, and it texts the caller nothing more. You get a Slack alert with this draft. With a hearing that afternoon, calling back beats texting, and if you're in court all day, someone at the firm needs to be watching those alerts.
A reply that asks the real question
The agent sends nothing here. A substantive question is flagged to you in a Slack alert with this reply drafted, and you decide whether to send it or pick up the phone. As a later text in the thread, it skips the firm name and the opt-out line.
The templates page has another law-firm opener next to other trades'.
Is an automated text to a caller a solicitation?
Here is what the ABA's model text says. It isn't advice, and it isn't the rule you practice under: each state adopts its own version, and some differ from the model in ways that matter for advertising and client contact.
Model Rule 7.3, solicitation
Rule 7.3 defines solicitation as a communication initiated by or on behalf of a lawyer, directed to a specific person the lawyer knows or reasonably should know needs legal services in a particular matter, that offers or can reasonably be understood to offer those services. The comment adds two points that bear on this agent. A communication responding to a request for information is not a solicitation. And the restriction on live person-to-person contact doesn't cover text messages. The agent is built around that distinction: it only texts someone who phoned first. Whether a missed call counts as a request for information under your state's version of the rule is a question for your ethics counsel. Rule 7.3(c) also bars soliciting anyone who has made known a desire not to be solicited, which is roughly what a STOP reply enforces.
Model Rule 1.18, prospective clients
Under Rule 1.18, a person who consults a lawyer about possibly forming a client-lawyer relationship is a prospective client. What you learn from them is protected even if you never take the matter, and learning enough of it can bar you, and your firm, from representing someone on the other side of the same or a substantially related matter. That is why these texts ask for nothing about the matter. The conflict check happens with a person, before anyone describes the facts.
ABA Formal Opinion 512, generative AI
Issued July 29, 2024, Formal Opinion 512 says lawyers using generative AI tools must consider their existing duties, including protecting client information and supervising the people and tools that work for them. It notes that Rule 1.18 extends confidentiality protection to prospective clients. This agent is a generative AI tool working on your behalf, so the opinion is worth reading before you sign anything, including with me.
The texting side works like any other business: the agent only replies to people who called, sends nothing between 8 pm and 8 am in your time zone, stops for good on STOP, and blocks promotional language. The number runs under Twilio's messaging policy. Take the whole setup to your state bar's ethics resources or your own ethics counsel before it goes live.
The caller's replies are read by a model. Know that going in.
No voicemail is recorded in the standard setup: an unanswered call gets a short spoken apology that says a text is on its way, and nothing the caller says is kept. What they text back is another matter. The agent reads each reply to write the next one, so whatever a caller types, including facts about their matter, passes through the AI provider that powers the agent. The first text asks people not to send details, but it can't stop someone from typing them anyway.
Conversation records are kept for 90 days. Every call and text is logged to HubSpot, or to any intake tool that accepts a webhook; alerts reach you in Slack. The agent doesn't write into Clio or any practice-management system directly. It attaches by call forwarding from your cell and by linking the consultation page you already use. If your confidentiality analysis says what a prospective client types can't go to a third-party model, stop here. That is a real reason not to buy.
The money is rarely the reason to say no.
$249 a month, plus AI and phone usage, typically $20–40 a month, after a $850 setup. Month to month, no free trial, and the number ports out with you.
Try it with example numbers. You miss 3 intake calls a week while in court: about 13 a month. If one caller in ten would have retained you at a $3,000 flat fee, that's 1.3 retainers, or about $3,900 a month, against roughly $289 for the agent. Your practice area moves those numbers more than anything on this page.
If what you want is a live voice on every call, look at an answering service built for that. As of September 2026, Smith.ai lists its AI receptionist as free for 25 calls a month ($3 a call after), with paid plans at $150 and $500 a month, month to month, on its pricing page. A voice answer suits callers in crisis better than a text.
Pass on this if
- Your practice lives on night calls. A call after 8 pm is never texted, that night or the next morning, and waits in your alerts.
- You have a receptionist or answering service already catching intake calls in business hours.
- Your work is referral-only transactional practice, where new clients rarely cold-call.
- Your confidentiality review rules out a prospective client's texts passing through an AI provider.
The setup steps, the review pass, and the carrier verification (about three business days) are on the missed-call text back page. Whether a text wins back a caller at all is weighed in does it work.