Front desk · clinic phone systems

Her call rang out.
Ten seconds later, she had a text.

A clinic misses calls because the people who answer them are with patients. The call is not lost because nobody cares — it is lost because both hands are busy.

Built and tested end to end on our own line. What follows is the system as it is built, workflow by workflow, including the one that ships switched off.

How it works Book a 15-minute call Your numbers, on the call
Missed call to text back, under ten seconds A patient calls while both hands are busy. The call rings out — drawn as a refusal — and the line picks it up: a text back to the caller in under ten seconds, then a row written to the leads sheet. A patient callsboth hands are already busyRINGS OUT — nobody is free to answerThe line picks it upunder ten secondsText back to the callera real reply, in her handLeads row written
The call still rings out. What changes is what happens next.

Every figure here is a build fact — counted, not claimed. No outcome numbers appear on this site until a client’s system produces one.

  • <10sfrom missed call to text back
  • 6workflows in the template
  • 5that run · 1 that catches the others failing
  • 1node changed per clinic
  • 7tests that pass before a line goes live
  • 7 daysfollow-up, stopped by any reply
  • 2hafter an appointment, the review request
  • 0phone numbers the clinic has to reprint

Five workflows that run, and one that catches them failing.

Five missed calls a day is ordinary for a busy front desk. Each one is somebody who was ready to book at the moment they were ready. By the evening she has booked somewhere else, and nothing in the clinic records that it happened — which is why the number never comes up in a staff meeting.

So the line catches what the desk cannot. A missed call triggers a text back in under ten seconds. After hours, a receptionist answers, qualifies and books. A new enquiry gets a seven-day follow-up that stops the moment they reply. A finished appointment triggers a review request. The clinic keeps its published number — it forwards to ours on no-answer, which is the only arrangement a clinic should accept.

How it works

Six workflows. Five run; one waits for the others to fail.

Numbered the way they are numbered in the template, because that is how they are cloned.

  1. 01

    Missed-call text-back under 10s

    The clinic phone rings first. If nobody picks up, the caller gets a text in under ten seconds and a row appears in the leads sheet. No AI in this one at all — it is the highest-return workflow in the set and the simplest thing here.

  2. 02

    AI receptionist after hours

    Answers outside opening hours, qualifies, and books into the real calendar. It may state only the services and policies written into its config — asked something it does not know, it offers a callback rather than guessing. That is the single most important test in the go-live checklist.

  3. 03

    Lead follow-up seven days

    A scheduler, not a week-long wait. The sequence stops the moment they reply and honours STOP the moment it arrives. Holding an execution open for a week means one restart drops every lead mid-sequence.

  4. 04

    Review engine two hours after

    Fires once, two hours after a completed appointment. A sheet of who has already been asked is the only thing preventing a second request, which is why nothing is ever deleted from it.

  5. 05

    Reactivation ships off

    A single rebooking offer to patients dormant ninety days — and the one workflow that ships switched off. It is marketing to a stored list, so it stays off until the clinic confirms in writing that the list opted in.

  6. 06

    Error handler catch-all

    Every one of the five points here on failure. It emails us and writes a row to a shared failures sheet, so a broken clinic line is something we find out about rather than something the clinic discovers.

The workflow nobody writes

The sixth exists because the other five will fail eventually. A phone system that fails quietly is worse than no phone system — the clinic stops checking, and by the time anyone notices, a fortnight of calls has gone nowhere.

Six workflows: five that run, one that catches Missed-call text-back, the after-hours receptionist, the seven-day follow-up, the review engine and reactivation — reactivation shipped switched off. All five route to a sixth workflow, the error handler, which emails us and writes a failures row. FIVE THAT RUN, ONE THAT CATCHES01 Missed-call text-backunder ten seconds, no AI02 AI receptionistafter hours — qualifies, books03 Lead follow-upseven days, stops on reply04 Review enginetwo hours after the appointment05 Reactivationships switched offWHEN ANY OF THE FIVE FAILS00 Error handleremails us, writes a failures rowA phone system that fails quietly is worse thanno phone system — the clinic stops checking.
The sixth is the one nobody writes, and the reason the other five can be trusted.

What you get

Configuration, not construction.

One node

Client Config

In every workflow there is exactly one node to edit: numbers, hours, calendar, clinic name, timezone. Two workflows carry a second copy because the phone provider calls those webhooks cold — miss one and a clinic answers with another clinic’s prices.

Seven tests

Before every go-live

Call it in hours. Let it ring out and watch the clock. Call it after hours. Ask it something it does not know. Book end to end. Reply STOP. Then break a credential on purpose and confirm the alert arrives.

Their zone

Never the server’s

Every hour is computed in the clinic’s timezone. Quiet hours, opening hours and the follow-up schedule read from the same config field, so a server on another continent cannot text a patient at 3am.

We do not go live on an untested path. However tight the date, all seven tests pass before a line is activated.

Want this one walked through against your own numbers? Email info@adapton.io and we will show you the parts that transfer and the parts that do not.

The limits

The stops, and the one we will not sell around.

Messaging law is a real risk to the clinic first and to us second, so the limits are built in rather than remembered.

  • STOP flips the lead to stopped and nothing further sends. Permanently, not for a week.
  • Quiet hours are enforced on every outbound message, in the clinic’s own timezone.
  • Reactivation runs with a hard per-run cap that is never raised to catch up — one malformed date column would otherwise text an entire patient list.
  • Reactivation stays switched off until the clinic confirms in writing that their list opted in to marketing texts. Statutory damages run $500–$1,500 per message.
  • We guarantee the system runs and that every missed call gets a response. Never a booking count — nobody can promise how many of the people it reaches will say yes.
Four things that refuse STOP flips a lead permanently. Quiet hours block outbound messages. A per-run cap is never raised. Reactivation stays off until consent is confirmed in writing. All four end at a refusal line. FOUR THINGS THAT REFUSESTOP, anywherestatus flips, and it is permanentQuiet hoursnothing outbound outside themA per-run capnever raised to catch upReactivationoff until consent is in writingStatutory damages run $500–$1,500 per message.
Each of these is in the build, not in a document somebody remembers.

Next step

Show us the thing you do every morning.

Fifteen minutes. We ask three questions and do the arithmetic with your real numbers. If it is not worth automating we will say so on the call.

Send us the details

All case studies · info@adapton.io