Train your receptionist
Teach it once. It answers forever.
A great receptionist knows your clinic cold — the prep for a scan, what insurance you take, where to park. Clinic Agent learns all of it from you, then answers patients in your words. It never invents an answer; every fact it uses is one you approved.
- Guided interview
- Bring your own documents
- Learns from real chats
- You approve everything
An agent is only as good as what it knows about your clinic.
Out of the box the agent handles the universal front-desk work — booking, rescheduling, hours. But the questions that actually stump a stand-in receptionist are the specific ones: “Do I fast before this blood test?”, “Do you take my insurance?”, “Can I bring my child?”. Training is how those answers get in — and how the agent keeps them accurate over time. Everything the agent knows lives in one tenant-scoped knowledge base, and only facts you’ve approved are ever allowed to reach a patient.
You teach it
A short back-and-forth interview, or a document drop — the agent turns what you say into clean, reusable answers.
It fills its own gaps
When a real patient asks something it can’t answer, it asks you — once — and remembers your reply for next time.
You stay in control
Nothing goes live until you approve it. Conflicting facts are held back for you to resolve, not silently overwritten.
The interview
Train it like you’d train a new hire.
Open a training session and the agent interviews you — one question at a time, the way a manager onboards a new receptionist. It asks a follow-up, restates what it heard so you can catch mistakes, and only saves the fact once you confirm it.
It keeps digging until it could answer a real patient without guessing, then stops. In a hurry? Turn on auto-confirm and clear answers are saved as you go. Every session is saved — rename it, pin it, or pick it back up later, with each fact traced to where it came from.
Bring what you already have
Drop in a PDF, a price list, or a link.
You’ve already written most of this down. Upload a PDF, CSV, text file or an image, or paste a link to a page on your site — the agent reads it, pulls out the answers, and walks you through confirming them. It never bulk-imports unverified facts; you still get the final say on each one.
Photos of a printed sheet are read directly when your configured model supports vision on a compliant provider; otherwise the agent simply asks you to describe what’s in the image. Either way the fact is yours to approve. Uploads are capped at 10 MB and links are fetched behind an SSRF guard.
insurance-accepted.pdf
PDF · 6 answers found
service-price-list.csv
CSV · 22 rows
parking-map.png
Image · describe or read
It asks when it doesn’t know
A question it can’t answer becomes a question for you.
When a patient asks something factual the agent hasn’t been taught, it doesn’t guess and it doesn’t dead-end. It tells the patient it’ll check with the team and quietly raises a “question from your receptionist” in your escalations.
You answer once. The reply goes back to the patient and — unless you say otherwise — is saved as approved knowledge, so the agent answers it directly the next time anyone asks. Repeat questions are de-duplicated so you’re never pinged twice for the same thing.
“Do you have wheelchair access at the main entrance?”
Your answer
Yes — step-free ramp on the left of the main doors.
It learns from real conversations
The questions patients keep asking become suggestions.
Every night the agent reviews where it fell short. When the same question comes up again and again — three times or more across the last month — and none of your approved answers cover it, it surfaces a suggestion: “patients keep asking this; want to answer it?”
It never writes the answer for you and never publishes on its own — the examples are stripped of patient details, and you either type the answer (which becomes approved knowledge) or dismiss it, in which case it won’t be proposed again.
Do you offer teeth whitening?
asked 7×Is there parking after 6 pm?
asked 4×Can I pay in instalments?
asked 3×
You stay in control
Every fact has a status, a source, and a shelf life.
Knowledge isn’t a black box. You can see, edit, approve, or retire any answer at any time — and the agent keeps it honest for you.
Approve before it’s live
Facts sit as draft until approved, and only approved facts ever reach a patient. Retire one and it’s instantly out of play.
Conflicts get caught
A new fact that contradicts an existing answer — or your Settings, like hours or services — is held back as a resolution card, never silently overwritten.
Nothing goes stale
Answers older than 90 days are flagged “needs re-verifying,” so a price or policy that changed doesn’t quietly go wrong.
See what earns its keep
Each answer shows how many times it’s actually helped a patient, so you know which knowledge is pulling its weight.
Traceable sources
Every fact is labelled by where it came from — interview, a patient question, a suggestion, onboarding, or an import.
Edit it like a doc
Manage the whole knowledge base directly from Settings — write, revise, or replace any answer without a training session.
Organised around how a clinic actually works.
Knowledge is grouped into the topics a front desk fields every day. Three of them — hours, services and pricing, and your doctor roster — are already pulled from your dashboard Settings, so the interview doesn’t waste your time re-asking them.
- Clinic basics & hoursfrom Settings
- Services & pricingfrom Settings
- Doctors & stafffrom Settings
- Scheduling policies
- Insurance & billing
- Prep & logistics
- Tone & communication
- Boundaries
How a trained answer reaches a patient.
Two paths, both grounded in your approved facts — and both sitting underneath the agent’s safety rules, never above them.
In the conversation, in context
Your most-used, most-relevant approved facts are given to the agent as it talks — ranked by usefulness and trimmed to fit, so the answers that matter most are always the ones it has on hand. They’re treated as reference data: they inform answers but can never override the guardrails that keep patients safe.
A reliable exact-match floor
For clear, direct questions there’s also a deterministic keyword match that answers straight from your approved knowledge without leaning on the model at all — a dependable floor beneath the conversational layer. (It’s keyword and usage based, not a fuzzy guess.)
Watch your receptionist get smarter.
The training dashboard shows facts learned over time, your coverage across every category, and a gap trend — how often the agent had to ask for help this week versus last. As the curve bends up and the gaps fall, you can see the front desk getting quieter.
Start trainingSee it handle your front desk.
Talk to the agent yourself in the live demo, then sign in to set it up for your clinic — most are live in about fifteen minutes.