On October 26, 2026, Clio opened ClioCon in Boston with a keynote that ran through so many product launches that Bob Ambrogi's LawSites wrap-up put "And It's A Lot" in its headline. Buried in the list was a feature most managing partners will care about more than drafting or eFiling. Clio's Grow AI "answers inbound calls, asks a firm's screening questions, runs conflict checks and books consultations," and it automatically follows up on incomplete intake forms and unsigned agreements. When Clio first launched Grow AI on August 20, 2026, it set the price at $25 per converted lead. If the lead doesn't convert, the firm pays nothing. In Clio's words, "If Grow AI had no role in the hiring, there's no charge."
I like that pricing. It is honest about what the product is for, and I would still not let it decide how my firm's intake works.
Say you are the managing partner of a firm with 8 to 40 people. You have probably heard the pitch for an AI agent that does the first pass of client intake follow-up: it chases missing facts, gathers conflict-check inputs, and sends engagement-letter reminders while your attorneys sleep. The pitch is right. The follow-up gap is real, it costs you matters every month, and an agent closes it cheaply. My argument is narrower and, I think, more useful. The agent you deploy will want something, the thing it wants gets set by whoever writes its success metric, and if you let a vendor's billing model write that metric, your intake process quietly turns into a sales funnel with a bar license stapled to it.
The leak sits in the second email
Most small firms do the first contact well enough. A person calls, someone answers, a consultation gets booked. Matters get lost in the days after that, in the stretch where the firm needs three more facts, the spelling of the other party's name, and a signature on a document the prospect has already opened twice.
Clio's own numbers on this are brutal. In the data cited at the Grow AI launch, 79% of prospective clients expect a reply within 24 hours, 67% of firms don't respond to email inquiries, and 48% of firms don't respond to phone inquiries. At ClioCon, the company added that 44% of client inquiries arrive after hours. Put those together and you get a picture every small-firm partner already knows from experience. The prospect writes at 9 p.m., hears nothing until the paralegal gets to the queue at 10 the next morning, and has already called the next firm on the list.
The follow-up is worse than the first reply, because nobody owns it. The intake specialist owns the first call and the attorney owns the consultation. The gap between them, where the questionnaire sits half-filled and the engagement letter sits unsigned in someone's inbox, belongs to everyone, so in practice nobody works it. In a 15-person firm the job usually lands on whoever feels guiltiest that week.
That orphaned stretch is where an agent belongs. The job is repetitive, it runs on a schedule, it lives in writing, and it needs no legal judgment at all if you design it correctly. It is the cleanest case for automation inside a law firm I know of, and also the easiest one to design badly.
What an AI agent for law firm client intake follow-up should actually do
Strip the job down and it has three parts. Each one has a version that helps the attorney and a version that quietly does the attorney's thinking.
Missing facts
The agent's first job is to compare what the firm has against what the attorney said they need, and to ask for the difference. A family law intake might need dates of marriage and separation, the county, whether children are involved, and whether any prior orders exist. An employment intake might need the termination date, the employer's legal name, and whether the person signed a severance agreement. All of these are facts, and none of them are opinions.
Its prompts should sound like a careful paralegal. "We still need the date you received the termination letter. Could you send it, or a photo of the letter?" That is a request. The bad version sounds like this: "To evaluate the strength of your claim, we need the date you received the termination letter." That sentence tells a prospect the firm is weighing their case, and the prospect now believes something about their legal position that no lawyer has said.
The checklist of required facts has to come from your attorneys, written per practice area, and it should stay boring. The agent asks for the listed items, confirms receipt, and stops. If the prospect volunteers a story about what happened, the agent saves it word for word for the attorney and does not summarize it into a theory of the case.
Conflict-check inputs
Clio says Grow AI "runs conflict checks." I would split that phrase in two. Collecting the inputs for a conflict check is clerical work and an agent does it well. It gets the full legal names of adverse parties, former names, related companies, spouses, the insurer, the opposing counsel if one exists. It asks for spellings twice. It catches that "Mike's Auto" is probably "Michael Rossi Automotive LLC" and asks the prospect to confirm.
Running the search against your matter database is also mechanical, and the agent can do it. Deciding what a hit means is a lawyer's call. A name match on a former client from 2019 could be nothing, a waivable issue, or a reason to decline. That call belongs to an attorney, and the agent's output should be a flag with the evidence attached ("possible match: Rossi, former client, closed matter, 2019"), routed to a named person with a deadline.
Firms skimp on this step because the inputs feel trivial, and then the conflict system works on bad data. Most conflict misses I have seen started as a misspelled name or a parent company nobody asked about. An agent that never gets bored of asking "Does the company have any other names it does business under?" will improve your conflicts process more than any new search software will.
Engagement-letter reminders
This is the part Clio's pricing pays for, and the part with the most temptation in it. The agent sends the engagement letter after an attorney approves it, then reminds the prospect to sign. That's all it does here. It does not renegotiate the fee, explain what the scope "really" covers, or tell the prospect the firm can start immediately once the letter is signed unless an attorney wrote those words into an approved template.
The reminder cadence should come from the firm. Say you choose a reminder at 48 hours, another at five days, and a handoff to a human at ten. That schedule is yours to set, and the agent's only creativity should be in timing (it knows the prospect opened the email at 11 p.m., so it sends the reminder at 8 the next morning). When the prospect replies with a question about the letter's contents, the agent stops and routes the question to the responsible attorney.
The line the agent never crosses
Here is the rule I would put in every firm's intake policy, in a sentence a nonlawyer can enforce. The agent never tells anyone, directly or by implication, whether they have a case.
Oregon gave the clearest regulatory read on this so far. Oregon State Bar Formal Opinion 2026-208, released in February 2026, asked if lawyers may use autonomous AI agents to deal with prospective clients during intake. Its answer, as reported by the Washington State Bar's NWSidebar, was "Yes, qualified." The qualifications matter more than the yes. Lawyers must understand how the agent responds to particular prompts and monitor what it does, so it does not appear to create an attorney-client relationship, promise services the firm cannot deliver, or guarantee results. The opinion says autonomous agents "warrants close oversight by the lawyers involved," and treats the duty under RPC 5.3 as supervising the service provider. Disclaimers, it adds, may handle some of the risk and won't necessarily handle all of it. Its closing line is the one to tape to the monitor: "Ultimately, Oregon lawyers are professionally responsible to supervise the accuracy of all work product."
The case assessment line is harder to hold than it sounds, because of who is now on the other end of the conversation. Clio's 2026 Legal Trends Report, released at ClioCon, found that nearly one in four U.S. adults has used AI to answer a legal question, and 46% would consult AI before contacting a lawyer. Your prospects arrive pre-briefed by a chatbot that has already told them, confidently, that they have a strong wrongful termination claim. They will ask your agent to confirm it, again and again, and in friendly words.
"So you think I have a good case?"
An agent tuned to be helpful will want to say something warm. An agent tuned to convert will want to say something encouraging. Both are dangerous. The only safe answer is a fixed one, written by an attorney, that the agent cannot paraphrase: something like "Only an attorney at the firm can assess your situation, and they will do that in your consultation. I can help get your information ready for them."
The same report found 82% of people want to know when their lawyer uses AI. That number settles the disclosure question for me. Tell prospects plainly that the follow-up messages come from an automated assistant that the firm supervises. Hiding it buys you nothing and risks the trust you are trying to earn before the engagement letter is even signed.
The meter and the mandate
Back to the $25.
Per-conversion pricing is a sensible product decision for Clio, and for many firms it will be a good deal. A firm with a $5,000 average matter will happily pay $25 for a client it would otherwise have lost to a slow inbox. Clio also lets firms set a monthly spending target, and existing Grow customers get their first three conversions free. None of my concern is about the price.
My concern is what the price teaches the system to want. At launch, Clio described Grow AI's lead scoring this way: "Every hired matter, every declined case, and every past outcome becomes a signal." Read that sentence as a managing partner. The system learns which prospects your firm tends to take, and it uses that learning to decide which leads to book and how hard to chase. Clio's chief product officer, John Foreman, framed the launch around access: "A two person firm shouldn't need a call center or a marketing department to compete." That is a fair goal. It is also a description of a sales operation, and sales operations optimize for yes.
A lead score built on past hiring decisions is a statistical guess about who has a case worth taking, which is the very judgment the agent must never make. I am not accusing Clio of anything. The firm configures the criteria, and Clio's ClioCon materials talk about firm-defined "playbooks" that mark where AI must stop for lawyer judgment. I am saying the incentive points one way, and a small firm without a written intake policy will drift with it.
The drift looks like this in practice. The agent chases the prospects it scores high and lets the low scorers age out of the queue. Nobody decided to decline those people. They just never got their second email. Six months later the firm's intake looks efficient, conversion is up, and a set of prospective clients (some with real claims, some with time-sensitive deadlines) were filtered out by a model trained on what the firm used to like. Under your duties to prospective clients, "the software didn't follow up" is a weak answer, and Oregon's opinion makes the supervising lawyer the one who gives it.
My position is blunt. For the follow-up stage, measure the agent on completeness and response time, and keep conversion off its scorecard. Lead scores, if you use them, go to the attorney as one input among several. The agent sends every prospect the same follow-up sequence until a human declines or accepts the matter. A declined matter gets a non-engagement letter that an attorney approved, sent on a clock, every time.
The SOP is the spec
The most interesting ClioCon announcement for small firms may have been the least flashy. Clio's new Agentic Workflows let firms "describe a process in plain language, or upload an existing standard operating procedure," and Clio builds the steps. Firms can inspect and adjust those steps and add approval points before turning the workflow on.
That design shows you where the real work moved. The software is the easy part now. The hard part is a written intake procedure, and most firms with 8 to 40 people do not have one. They have a paralegal who knows how intake works, a partner who has opinions about it, and a shared inbox. When you upload an SOP to an agent, every vague sentence in it becomes a decision the model makes for you.
Write it the way a new hire would need it
For illustration, here is the level of specificity that works. "For personal injury intakes, request the date of the incident, the location, the names of all drivers, the insurer for each vehicle, and whether a police report exists. If the incident date is more than 18 months ago, stop the sequence and notify the supervising attorney the same day." The last clause matters most, because limitations periods are where a chatty agent can do the most damage by sending a cheerful reminder about next week's consultation. A rule that turns a date into an escalation keeps the deadline judgment with the lawyer and the arithmetic with the machine.
Then list what the agent may never say: predictions about outcomes, fee estimates outside the approved letter, "we can help you with that," and any reassurance about deadlines. Write them down as forbidden phrases and test them by trying to make the agent say each one, in a dozen polite ways, before it ever touches a real prospect.
Put a name on every approval point
Clio lets you add approval points. Each one should have a named person attached and a time limit, and a fallback when that person is out. "Attorney review" is not a step. "Dana Li reviews conflict flags by noon the next business day, and Sam Ortiz covers when Dana is out" is a step. An agent that routes a conflict hit into a queue nobody watches has made your process slower, while the dashboard shows it running perfectly.
Read the transcripts every week
Oregon asks lawyers to monitor what the agent does. In a small firm that means a partner reads a sample of real follow-up threads every week, looking for three things in particular: drift toward case assessment, prospects who went quiet without a human ever looking, and conflict inputs that came back incomplete. Say it takes a partner 45 minutes a week. That is the cheapest malpractice insurance the firm will ever buy, and it gives the partner something no vendor report will, a direct feel for what prospects are worried about before they ever sit down in the conference room.
Why the smallest firms gain the most
Clio's report says AI adoption reached 85% of firms, up from 79%, and projected lawyer utilization rose to 42% of an eight-hour day, from 39% in 2025, which the report called "the largest single-year increase in the report's history." Large firms have intake departments, business development staff, and conflicts teams with their own managers. A 12-person firm has a receptionist and good intentions. The follow-up agent gives the small firm something the big firm already pays humans to do, which is a reliable second, third and fourth touch.
That advantage only holds if the small firm keeps control of the rules. A big firm's intake team has written procedures because it had to train people. A small firm that skips the procedure and switches on a vendor default has handed its intake standards to whatever the default optimizes. The firms that write the procedure first will end up with a cleaner conflicts database, faster engagement letters, and fewer prospects who fall through the cracks. Their attorneys will also spend their consultations on legal questions, because the facts arrived complete.
Architect it before you buy it
The tools are here. Clio shipped one in August and expanded it in October, and there will be five competitors by spring. Buying the tool takes an afternoon. Designing what it may want, what it may say, who it answers to, and when it has to stop is the work that decides whether intake automation protects your license or slowly tests it. That design has to fit your practice areas, your conflicts system, your engagement letters and your partners' actual tolerance for risk, and no vendor default was written with your firm in mind.
Agor AI Advisory builds that layer with managing partners. We write the intake SOP with your attorneys, map the forbidden phrases and escalation rules, wire approval points to named people, and set up the weekly review so the supervision Oregon describes actually happens. Then we help you pick and configure the tool against that spec, so the agent works for the firm's judgment instead of around it.
Your next prospect writes in tonight at 9 p.m. Decide now what the machine is allowed to want when it answers. Schedule a strategic consultation with us today.
Sources
- LawSites (LawNext), ClioCon 2026: A Quick Wrap-Up of Everything Clio Announced Today, October 26, 2026
- LawSites (LawNext), Clio Grows Clio Grow with Launch of Grow AI, Providing 24/7 Intake Agents, August 20, 2026
- NWSidebar (Washington State Bar Association), New Oregon State Bar Opinion on Using AI Agents for Client Intake, March 23, 2026