AI automation for medical clinics on the front desk, not in the consulting room
The front desk picks up the phone, checks the schedule, writes down a slot, calls back to confirm and explains for the hundredth time how to prepare for a test. We take that part over, the administration of the practice. The AI agent does not diagnose, does not advise on treatment, does not read test results and does not touch patient records.
- 14 daysto the first working automation, counted from the end of the workshop
- EUR 3,500 netper week of workshop, a published price we do not negotiate
- 0 entriesthat is how much the agent writes into a patient record. It does not read it, does not change it and does not judge results
The problem
Why does the front desk spend half a day on the phone?
A patient calls to book, to move an appointment or to ask whether they should come in fasting. The front desk picks up, checks the schedule, writes the slot down and calls back to confirm. The same conversation repeats dozens of times a day, and a patient who cannot get through in three minutes books with the practice next door.
A cancelled appointment leaves an empty slot. Somebody has to notice it, pull up the waiting list and call before the time is lost. With a dozen specialists on the schedule nobody does that as it happens, so the slot usually just goes, while the cost of the room and the staff stays exactly the same.
Then there is a layer that other industries do not have. Patient data is a special category of personal data under Article 9 of the GDPR, so every tool that touches the booking process has to be described, limited and covered by a data processing agreement. That is why in a medical practice we start with the boundary of the rollout, not with a list of features.
- A patient waits on hold and books with the practice next door
- The slot left by a cancellation is lost, because nobody called the waiting list in time
- The front desk explains opening hours, parking and test preparation for the hundredth time
- Consent forms and the intake questionnaire are filled in at the desk, out of the appointment time
Why now
What changed, so this can be done without touching patient records?
Practice management and scheduling systems now have interfaces that let you read free slots and write a booking back without replacing the software. Three years ago wiring that together was a quarter-long project. Today it is counted in working days.
The second change matters more. A rule will move a booking from the system into a calendar, but it will not answer a patient asking whether they have to fast before a test. A language model answers in a sentence, from the knowledge base of the practice, at any hour. What counts is what it does not do: questions about symptoms, results and treatment go to the staff.
The third thing is organisational. Work can now be split so that the AI agent sees only the calendar layer, the contact details and the billing data, while the patient record stays out of its reach. That is not a setting in a panel, it is a design decision we take in the workshop and write down.
What changes
What does a practice have after the first month?
Booking without waiting on the line
A patient books a slot through your website and gets a confirmation in the same minute. The phone stays for those who prefer to call, it just stops being the only way in.
Confirmation by SMS and email right after booking, with no call back
An empty slot goes back into circulation
A cancellation triggers the waiting list for that time, instead of waiting for somebody to look at the schedule. The offer goes to patients in the order your practice sets.
A cancellation at 21:00 starts the queue at once, not the next morning
A reminder a patient can confirm with one tap
The patient gets the time and the preparation instructions, and answers with one tap. The front desk stops phoning down the schedule the day before.
Two reminders: 48 hours and 2 hours before the appointment
Organisational questions have an answer at any hour
Opening hours, how to get there, how to prepare for a test, the price list, the documents to bring. The answer comes from the knowledge base of the practice, not from the general knowledge of a model.
A question about symptoms or treatment goes to the staff, the agent does not answer it
Consents and intake forms before the visit, not at the desk
The patient fills in the organisational forms at home and the staff file them in the record. The appointment starts with a conversation instead of paperwork.
The form goes out with the booking confirmation, complete before the visit
Invoices and room utilisation counted
The sales document is built from the booking data, with nothing retyped. A utilisation report shows which rooms and which hours are standing empty.
An invoice in your billing system, a utilisation report every week
Proof
Do you use this yourselves?
We do not have a rollout in a medical practice whose numbers we are allowed to publish, and we are not going to get around that with an anonymous example. So we show the only thing we can show in full: an AI agent running in production in our founders' own company, built with the same method. Together with what those numbers do NOT prove.
Built for ourselves. These are not client deployments but implementations inside the founders' own companies. We say so plainly, because proof you cannot verify is not proof.
JustBottheir agentJustIdea Agency, a performance marketing agency of around 30 people
- 23
- people from the company work with the agent
- 38
- systems wired in permanently
- 112
- processes run without a human
- 475
- thousand lines of the agent’s code
The agent is wired into the task system, the team calendars, the CRM and the data warehouse. First commit 1 March 2026, 23 people use it today. It answers organisational questions: who is available, what is happening with a case, what came in from a form.
8 of 38 systems it works in
Google Ads
Meta Ads- HubSpot
- ActiveCollab
- Ahrefs
GA4
Search Console
BigQuery
- Before
- Questions about who was free and where a case stood circulated between people. The answer depended on who happened to have the right system open and a spare moment.
- After
- The same questions go to the agent and come back from data, in seconds. A person steps in where something has to be decided, not where something has to be looked up.
8,631 operations on team availability and 1,301 on form enquiries in 30 days
And in your practice
What if things work differently here?
Four sentences we hear in practices more often than any others, and all four are fair.
Data separately, because that question always comes back. We keep the data scope minimal, grant access per role, sign the data processing agreement before launch, and every message that goes out to a patient is approved by a person. That is not a legal opinion, it is how we design a rollout. We write about it at more length in our guide to AI and GDPR.
We have our own booking system and we are not going to change it.
There is no need to. We read free slots through the interface and the booking goes back into your system, which stays the source of truth for the schedule. Replacing a practice system is a quarter-long project, and here the point is fourteen days to the first result.
Most of our patients are older people, they will not book online.
That is why the phone stays. Automation takes the repetitive calls off the front desk so it has time for the ones that need a person. Reminders go by SMS, because SMS works on every phone and needs no app.
We are not giving an AI access to patient records.
We are not asking for it. The agent works on the calendar, the contact details needed to book, the organisational forms and the billing data. Test results, visit notes and treatment plans stay in your system, on the staff side, and we write that boundary down in the workshop.
Who is responsible if the agent writes something wrong to a patient?
Messages that go out to a patient are approved by a person, and the templates are agreed and written down before launch. The agent does not improvise on medical matters: a question about symptoms, dosage or a test result is passed to the staff instead of being answered.
- Booking and scheduling
- practice management and scheduling systems, room and equipment calendars
- Patient contact
- phone, SMS, email, the form on your website
- Teamwork
- Microsoft 365 or Google Workspace
- Billing and invoices
- your invoicing system, including an on-premise program reached through a connector inside your network
- Reporting
- spreadsheets, a room utilisation report, summaries for accounting
- Automation layer
- Make, n8n, custom API integrations
What we take over and what we never touch
The split of work in a practice before a rollout and after it. The last line matters as much as the rest.
| Area | Today | After the rollout |
|---|---|---|
| Booking an appointment | a call to the front desk during opening hours | online booking and the phone, confirmed in the same minute |
| Cancellation and empty slot | noticed when somebody looks at the schedule | a cancellation starts the waiting list at once |
| Reminders | the front desk phones down the schedule the day before | SMS and email 48 hours and 2 hours before the visit |
| Organisational questions | a queue on the phone line during opening hours | an answer from the knowledge base of the practice, at any hour |
| Consents and intake form | paper filled in at the desk | an online form before the visit, filed in the record by staff |
| Invoices and billing | retyped by hand from the booking | a document built from booking data in your billing system |
| Patient records, results, diagnosis | the doctor and the staff | the doctor and the staff, unchanged, with no agent access |
The scope of the first rollout is agreed in the workshop. We take the three lines with the highest return, not all of them at once. The last line we never take.
How we work
Five steps from the phone to a schedule that looks after itself
Every step ends with something you can judge before you decide on the next one.
Workshop and the boundary of the rollout
One week. We walk the patient route from the first call to the invoice and mark the boundary on it: what is administration and what is medical work. Everything on the medical side stays with the staff, and we put that on paper before we build anything.
Data scope and the processing agreement
We write down which fields are needed to book, remind and bill, who has access to them and where they sit. The data processing agreement is signed before launch, not after it.
First automation live
Fourteen days to a working solution. In a practice we usually start with confirmations and reminders, because those take the largest share of repetitive calls off the front desk.
Pilot on one specialty
We run it on a single specialty, alongside the current way of working. For the first days every message to a patient is approved by a person, until the templates are settled.
Production and measurement
After thirty and ninety days we check the indicators agreed in the workshop: how many slots went back into circulation, how many calls came off the front desk, how many invoices were raised without retyping. Only then do we decide on the next area.
Before and after
How a day on the front desk changes
Before
- The phone rings while a patient is waiting at the desk, and you have to choose
- A cancelled appointment stays an empty gap in a specialist's schedule
- A patient arrives unprepared, because nobody had time to call
- Room utilisation is known after the month closes, from a spreadsheet
After
- Repetitive questions have an answer and the front desk takes the hard ones
- A cancellation starts the waiting list for that slot immediately
- The reminder with instructions goes out by itself, twice before the visit
- Room utilisation shows up in a report every week
Who treats patients does not change. What changes is how much time the staff give to work that needs no medical knowledge.
For the practice owner and the front desk manager
What the person signing this off gets
The first thing is visible straight away: the front desk stops being a call centre and goes back to looking after the patients in front of it. Empty slots stop being a cost nobody hears about.
The second is quieter, and it counts more in an audit and in a conversation with your data protection officer. The data scope, access per role and the processing agreement are written down before launch, so the answer to who has access to patient data is a document rather than a reconstruction from memory.
- The boundary of the rollout written down: the agent does not touch patient records
- Data scope and access per role agreed before the first launch
- The data processing agreement signed before the start, not after it
- Indicators to measure after 30 and 90 days, agreed in the workshop
What it costs
The workshop price is published, the rollout price comes after it
The workshop costs EUR 3,500 net per week and we publish that before any call, because a practice has a right to know what it is signing up for. One or two weeks is usually enough in a clinic. We quote the rollout after the workshop, once we know how many systems we connect to and what state they are in.
You pay for SMS, the language model and the tools directly to the vendor, with no commission from us. A model query costs the provider roughly EUR 0.005 to 0.07. If the data has to stay inside your own network, we say so in the workshop and design the rollout that way.
Process workshop
EUR 3,500 net
per week, in a practice usually one to two
A self-contained piece of work. It ends with a map of the patient route and a 30 day plan you can execute without us.
- A week of work inside the practice and in the systems
- A map of the patient route from the first call to the invoice
- The boundary of the rollout written down: what stays with the staff
- Three automations with the highest return, counted
- The indicators to measure after 30 and 90 days
- Most popular
Workshop and first rollout
Quoted after the workshop
scope follows from the number of systems and specialties
For a practice that wants working confirmations and reminders within two weeks of the workshop ending.
- Everything in the workshop
- The first automation live in 14 days
- Connections to the booking system, email and the SMS gateway
- Data scope, access per role and the processing agreement before the start
- A pilot on one specialty, with messages approved by a person
- Measurement after 30 and 90 days
AI agent for a practice
Quoted after the workshop
for practices where organisational questions come in all day
For a practice where the front desk answers the same questions dozens of times a day and a dozen specialists share the schedule.
- Everything in the rollout
- An agent answering organisational questions from patients and staff
- A room utilisation report and alerts on empty slots
- Access per role for the front desk, management and accounting
If after the workshop we conclude that, beyond the medical records we do not touch, there is nothing here worth automating, we say so plainly and do not propose a rollout. The workshop still ends with a process map and a 30 day plan that holds without us.
Frequently asked questions
Does the AI agent diagnose or advise on treatment?
No, and this is the boundary we do not move. The agent works only on the administration of the practice: booking, the schedule, reminders, organisational forms, billing and the utilisation report. A question about symptoms, a test result or dosage is passed to the staff instead of being answered. We write that into the configuration and into the workshop document, so it does not rest on good will.
Does the agent read patient records?
No. It has no access to them and writes nothing into them. It sees the calendar layer, the contact details needed to book and confirm, and the billing data. Test results, visit notes and treatment plans stay in your system, on the doctor and staff side. The line between the two is drawn in the workshop and written down before anything is built.
What exactly do you automate in a medical practice?
Booking and confirming appointments, handling cancellations together with filling the freed slots, reminders before a visit, the stream of organisational questions about hours, directions, preparation, prices and documents, the flow of consents and forms, and finally billing, invoices and the room utilisation report. The scope of the first rollout is narrowed in the workshop to the three items with the highest return.
How do you handle GDPR with patient data?
Patient data is a special category of personal data under Article 9 of the GDPR, so we keep the scope minimal, grant access per role, sign the data processing agreement before launch and have a person approve every message that goes out to a patient. That is not a legal opinion, it is how we design a rollout; the guide to AI and GDPR covers it at more length. The compliance assessment on your side stays with your data protection officer or your lawyer.
Do we have to change our booking system?
No. We read free slots from it through the interface, the booking goes back into it and it stays the source of truth for the schedule. Replacing a booking system is a quarter-long project, while the first automation is meant to be running in fourteen days.
Our accounting program runs on a PC in the building, can that be wired in?
Yes. An on-premise program is handled through a connector running inside your own network, which takes the booking data and raises the document locally. The data does not have to leave your network to do it. The same arrangement works for a cloud invoicing service, except that there the connector is not needed at all.
By how much will no-shows drop?
We will not give you a percentage, because we have no measurement to take it from, and a number copied from somebody else's brochure means nothing for your schedule. Instead we measure your starting point before the launch and compare it after 30 and 90 days. The mechanism is simple: two reminders with one-tap cancellation, and a waiting list that gets the freed slot immediately.
How long does it take from signing to the first result?
One week of workshop and fourteen days to the first rollout, counted from the end of it. Then a pilot on one specialty, and we measure the effect after thirty and ninety days, on indicators agreed before the start, so that nobody judges it on impressions.
Let us start with one patient route
The free consultation takes 30 minutes. We walk one of your routes, usually booking and the reminder around it, and tell you what can be taken off it without touching patient records. If the answer is nothing, we say so straight away. You write directly to Paweł, no reception and no tickets.
We run at most two rollouts at a time, so that someone is actually sitting on each of them.
Book a free consultationPS. If your front desk explained how to prepare for a test for the hundredth time this week, and a cancellation left a gap in the schedule, that is exactly the cost. We take the repetitive calls off and keep an eye on the schedule. Treatment stays with the doctors.