It’s 9:40 on a Tuesday night. A man who just got rear-ended on the interstate is sitting in an ER waiting room with his phone, calling the first personal-injury firm he found. Yours. Three rings, then voicemail. He hangs up before the beep finishes and taps the next result. By the time you check messages tomorrow, he’s already signed a fee agreement with someone else. That call was worth a contingency fee in the tens of thousands, and you never knew it happened.
That’s not a worst case. It’s the median case. A 2025 national audit that placed 1,200 calls to small and mid-sized U.S. law firms found 34.8% went completely unanswered during regular business hours, which the authors peg as a $109 billion-a-year revenue conversion problem for the legal industry. After hours, with nobody at the desk, the number isn’t 35%. It’s nearly everyone.
We’re gmware, a software development firm headquartered in Austin, TX with engineering centers in Bangalore and Mohali, India. We build AI voice agents onto businesses’ phone lines, including legal-intake agents that sit on a real firm’s number. This post is the part most legal-answering pitches skip: what an AI intake agent actually does on a call, where the conflict check and qualification fit, and the one section that matters most, when a live human is still the right person to pick up.
What happens to a call to a law firm
Why the first firm to pick up usually gets the client
Legal intake runs on speed, more than almost any other purchase. Someone who’s just been hurt, arrested, served, or fired is not shopping carefully. They’re scared and they want it handled, so they call down a list and stop at the first firm that answers like a human and tells them what happens next.
The numbers behind that instinct are stark. MIT’s lead-response research found that contacting a lead in the first five minutes makes you 100 times more likely to reach them than waiting until the 30-minute mark. A voicemail box doesn’t make contact in five minutes. It makes contact never, because 80% of the people who reach it hang up without leaving a message and dial the next name instead.
Here’s the part that stings for firms that pride themselves on calling back fast. By the next morning, the injury caller has already retained someone. Returning a voicemail at 8am loses to a competitor who answered at 9:40pm. The clock on a new matter doesn’t run on your office hours. It runs on the client’s panic, and the firm that’s there when the panic peaks signs the fee agreement.
What an unanswered intake call costs your firm
The math is uncomfortable once you actually run it, because a single retained matter in many practice areas is worth thousands, sometimes far more. So a missed call isn’t a missed message. It’s a missed case.
Use the same simple model we use with every business, adjusted for legal economics:
Qualified calls missed per week × average matter value × 4.33 weeks = monthly revenue exposed.
Say you’re a small PI or family-law shop. You miss 8 genuinely qualified intake calls a week (after-hours, lunch, on-another-line, the receptionist stepped away), and a retained matter averages $4,000 to you after the dust settles. That’s 8 × $4,000 × 4.33, roughly $138,000 a month of exposed revenue at full conversion. Nobody converts all of those. But you don’t need all of them. Sign one in five and you’ve recovered about $27,000 a month from calls that were rolling to a dead voicemail box.
The missed-intake model (illustrative)
Two honest caveats. Matter value swings everything, so a $1,500 traffic ticket and a seven-figure injury case can’t share a row. And not every missed call is a real client; some are wrong numbers, solicitors, or matters you’d decline. Use a sign rate you’d defend to your managing partner. Even a pessimistic one usually clears the cost of answering the phone. For the full break-even against what a service charges, we lay it out in our AI receptionist cost breakdown.
What an AI legal-intake agent does on the call
Strip the marketing and an AI answering service is a voice agent on your firm’s line. A call comes in at any hour. It answers on the first ring, greets the caller in your firm’s name, and does the structured intake work that a tired front desk does inconsistently and a voicemail box doesn’t do at all:
- Captures the caller and the matter. Name, callback number, and the matter type in the caller’s own words (“I got into a wreck,” “I’m getting divorced,” “my employer fired me after I reported them”), so you wake up to a clean record instead of a name and a shrug.
- Runs a first-pass qualification. It walks your intake questions for that practice area: date of the incident, was anyone cited, is there an existing attorney, is the matter inside your jurisdiction and statute window. Unqualified callers get a polite off-ramp; the real ones get fast-tracked.
- Flags a possible conflict. It captures the opposing party and key names and checks them against your conflicts list, then flags a potential hit for a human to clear before anything advances. It’s a first-pass flag at the door, not a final clearance.
- Books or routes. A qualified, conflict-clear consult lands on your calendar. A genuine emergency, the arrest call at 2am, gets routed to your on-call attorney with the context attached. Everything else gets a summary texted to intake for the morning.
The unfair advantages are the ones a staffed service can’t match on cost. It answers every call, so the 9:40pm injury caller hears a real response instead of a beep. It handles several calls at once, so a multi-car pileup that lights up your line with five callers in ten minutes doesn’t turn into hold music. And it costs the same at midnight as at noon.
One line we’ll repeat because it matters: the agent intakes and routes. It does not give legal advice, quote outcomes, or opine on a case’s merits. Anything drifting toward advice is its cue to route the caller to an attorney. If you want the full pipeline and the guardrails behind a voice agent, we cover it in how an AI voice agent handles a call, and the broader pattern in our AI agents for business operations guide.
When a live legal-intake human is still the better fit
Here’s the verdict we’ll defend, and it’s the one most AI-answering pitches won’t give you. For a real slice of legal calls, the human wins, and it isn’t close.
Route to a trained person when the call is high-empathy or high-stakes and the wrong word does real harm. A grieving family calling about a wrongful death. A domestic-violence survivor who needs to be heard, not processed. A frightened client facing arrest who needs steadying before any intake. A high-net-worth prospect who expects to speak with a partner, full stop. Those calls carry emotional and legal weight that a bounded voice agent shouldn’t carry alone, and when you only get a handful a week, the per-call cost of a human barely registers, so paying for judgment is the easy call.
A live service also wins on the nuanced intake where the facts don’t fit the script, the matter that’s technically out of your area but worth a careful referral, the caller who’s confused about what kind of lawyer they even need. A person reads that and adjusts. And if your firm’s entire brand is “you’ll always reach a real person here,” an AI front line undercuts the promise even when it works flawlessly. Keep it.
The honest split for most firms isn’t AI versus human. It’s AI for the repetitive intake volume, the captures, the qualifications, the “are you taking new clients?” calls, with a clean warm transfer to a human the moment a call needs a heartbeat. Done right, the AI is the filter that makes sure the calls reaching your on-call attorney are the ones that actually need them. For the full side-by-side on where each one wins, see AI receptionist vs human answering service.
How gmware builds the intake side
When the split makes sense, we build the AI intake agent to your firm, not off a template. That means the voice agent itself, the connection into your calendar and case-management or CRM system so it can actually book and log a matter, the conflicts-list check, the qualification logic for each practice area, and the routing rules for who gets the 2am call. An AI receptionist is an AI agent on a phone line, which is why this runs through our AI voice agents service and our AI agents and LLM integration practice: speech in, a bounded model deciding what to do, voice back, wired into your real systems with a clean handoff to your team and a hard line against giving advice.
We don’t publish a flat legal-answering price, because the cost tracks your call volume, your practice mix, and what you’re connecting it to. We also run production systems of our own: our Shield Suite product tracks retail intelligence across 60,000+ beverage-alcohol storefronts, so the reliability and escalation discipline behind an always-on phone agent isn’t theory. Delivery pairs Austin oversight with engineering in Bangalore and Mohali, which keeps the cost mid-market sized.
The full capability set sits on our AI receptionist hub if you want the wider view first. Tell us how many intake calls you think you’re losing and what your matter mix looks like. Reach out and we’ll give you a straight answer on scope, cost, and timeline within 48 hours, including whether an AI front desk is even worth it for your firm’s volume.