Phone Answering
The phone rings at 8:40 on a Saturday night. This agent picks up on the first ring, says up front what it is, and works the intake — name, callback number, service address, what's actually wrong, how urgent it is — reading each answer back to the caller before it moves on. Routine jobs get booked into an open slot. Anything outside the script stops and goes to a person.
Yours outright — no required subscription.
Every agent is built for your business — your systems, your approval chain, your way of working. We scope it on the call.
If this sounds too technical, don't worry. We take care of everything for you.
- Private VPS deployment. This agent runs on a server that belongs to you, not a shared cloud tenant. Your recordings and transcripts live there too, under your keys.
- A technician agent alongside it. Every hire ships as a pair: the Phone Answering Agent taking the calls, and a VPS-technician agent on the same box keeping it patched, backed up, and monitored. You're not hiring one thing, you're hiring a small team of two.
- Connection to the number you already publish. It sits in front of the line that's on your trucks and your cards, so nothing your customers have has to change. You decide what rings a human first and what it answers.
- Read-back confirmation on every field. Street name, callback number, spelling of the last name, day of the appointment — it repeats each one to the caller and gets a yes before moving on. Phone audio is the worst input there is, and this is what catches a wrong digit while the caller is still on the line.
- An escalation rule scoped with you. Explicit triggers — gas, smoke, water coming through a ceiling, chest pain — are set during setup to stop the intake and hand off to a person immediately. You name the triggers and you name who gets the call.
- Security hardening. Private VPN, firewall, and encryption configured on your server before it ever answers a live call. Recording disclosure is set to your state's rule at the same time.
- A live walkthrough, then 14 days of priority support — we play back real calls with you so nothing is a black box, then tune it against your actual call mix for two weeks. After that we remove our own access.
Not sure it fits? Check fit in 90 seconds in the free assessment chat.
Who this is for
- You publish one number and it rings the office manager's desk while she's checking in a tech and running a card. Line two rolls to voicemail, and you couldn't tell anyone how many of those callers never left a message.
- You already pay an after-hours service, and what shows up at 6am is a name, a callback number, and "no heat." Someone still calls back to get the address and the unit type, then re-keys the whole thing by hand.
- You cover nights and weekends with your cell phone, or your spouse's. You've taken a call about a leaking water heater from a soccer field, on a ladder, and at 11:40 at night.
- You run a clinic where the front desk is also doing check-ins and copays, so reschedule calls land in voicemail and get worked at lunch — after the slot the caller wanted was already given away.
How it earns trust
The failure mode of AI on the phone isn't that it mishears something — people mishear things too. It's that the call ends, the caller is gone, and there's no way to check what was actually said. This agent makes the opposite bet: every call leaves a record you can open and read.
The call is stamped onto whatever it created
Every booking, job record, and escalation text carries a link back to the call that produced it. When a tech rolls to the wrong address, you don't argue about it — you open the booking, click through, and hear the caller say the street.
The intake record shows the decision
Name, callback number, address, the problem in the caller's words, urgency, and how it classified them are stored as fields next to a timestamped transcript. When it books a slot you wouldn't have, you can find the sentence that made it decide.
A daily list of the misses
Every call it escalated, every caller who hung up mid-intake, and every call it couldn't classify land on one list each day. "How often does it fail" is a number you read, not a question you ask.
The whole archive — recordings, transcripts, intake records — lives on your server under your keys, not in a portal you lose access to when you stop paying.
Pairs well with
These share a workflow with this role. Tick any to add them to your setup.
We're missing calls all day and I couldn't tell you how many.
That's the part it fixes first. It answers every line at once, so nothing rolls to voicemail because someone was already on the phone, and every call it touches shows up in a log with a recording attached. That includes the ones that go badly: callers who hang up mid-intake are on the same list. Within a week you'll know the real number instead of guessing at it.
The answering service just takes a message. I still have to call them back to find out anything useful.
This one runs the intake instead of taking a message. It captures the callback number, the service address, what the caller says is wrong, and how urgent it sounds, reads each field back, and writes a structured record into your system of record rather than emailing you a paragraph. You still get the recording and the transcript. What you don't get is a 6am callback just to learn what kind of unit it is.
I don't want it lying to my customers, and I don't want it quoting prices.
It never quotes a price, offers a discount, promises an arrival window it can't verify against your board, or answers whether something is under warranty or covered by insurance. Those stay your conversations, and that boundary doesn't move. It also never takes a card number, a Social Security number, or an insurance ID over the phone — callers who want to pay get a person or a link. And it never places outbound calls: no reminder blasts, no callbacks, no campaigns. It answers the phone.
What happens if somebody calls with a real emergency?
The explicit triggers you scope with us — gas, smoke, fire, water coming through a ceiling, chest pain — stop the intake cold and route to a human immediately. Its job at that point is to hand off, not to help. Be clear about the limit though: it has no clinical or safety judgment, so a caller who says the house "smells kind of funny" or that she "just doesn't feel right" may get handled as routine. That's a real gap, not a setting we can tune away, which is why the escalation list gets written with you before it goes live.
By the time I call back, they've already booked somebody else.
For routine work it books the slot on the call, against the openings your board actually shows, and the caller hangs up with an appointment instead of a promise. Two honest limits. It won't commit to an arrival window it can't verify. And it's an intake worker, not a salesperson — a caller who is price-shopping and needs you on the phone will get a competent intake instead, and you'll be re-selling that one yourself.
What's covered under the $1,997 setup, and what isn't?
One company, one system of record, one approval chain, one on-call rotation. Typically about seven days from kickoff to it answering real calls. The thing to plan for after that is drift: add a service area, drop a brand you'll work on, change who's on call, and the script still reflects January. Nothing breaks loudly — it just starts booking jobs you don't take. Multiple locations or genuinely different call flows per line get scoped on the 15-minute call.
Hiring more than one? A department on tap — the subscription puts a build team behind every request, agent after agent.