Clinic owners usually ask the same question first: not “can it answer the phone?” but “will my patients hate it?”
The short answer is no — not simply because it is AI. Patients react badly when a system blocks a person, misunderstands the request, hides what it is, or drifts into medical judgement. Remove those four failures and the evidence becomes much more positive.
That does not mean every patient will welcome an AI receptionist. Some want a human, and some calls need one immediately. The useful question is whether the system makes a routine task easier without trapping the caller inside it.
What the patient research says
Two surveys tested comfort with AI in healthcare using different samples and different wording. Both put appointment scheduling among the most accepted uses.
KLAS Research and Luma Health surveyed just over 1,000 US patients. 69% said they were very or somewhat comfortable with AI handling appointment scheduling.[1] Sogolytics surveyed 1,012 adults and found a lower baseline, but scheduling still ranked first among the AI tasks tested.[2]
The most revealing number was not the baseline. It was what happened when access to a person disappeared.
Patients are not rejecting AI. They are rejecting AI used as a wall.
The four failures that make a call feel cheap
The research points to four distinct risks. Treating them as one vague fear makes the system harder to evaluate.
No route to a human
A fast automated answer stops feeling useful the moment the patient cannot leave it. Human handoff is not an edge feature. It is part of the main flow.
The caller is not understood
Repeating a surname three times or being forced into the wrong category makes the interaction feel designed for the clinic, not the patient.
The system pretends to be human
A disclosure sets the right expectation. Concealment turns an ordinary administrative call into a trust problem.
Admin turns into medical judgement
Booking a dentist is an administrative task. Deciding what a symptom means is not. The boundary belongs in the product, not in improvisation during a call.
A 2025 JMIR qualitative study found broad openness to AI support tools for administrative work. The barriers were lack of human oversight and the feeling that the system served the organisation rather than the person calling.[3]
That distinction matters. A patient may accept a two-minute automated booking and still object to a system deciding whether their pain is urgent. Those positions are consistent.
Where patients draw the line
A JAMA Network Open study surveyed 13,806 patients across 74 hospitals in 43 countries. 72.9% wanted a physician to make the final call in an AI-assisted diagnosis. Only 4.4% supported fully autonomous diagnostic AI.[4]
The implication for a clinic phone system is straightforward.
Disclose
Say clearly that the caller is speaking with an AI assistant.
Handle admin
Check real availability, book, reschedule, cancel, or capture a callback.
Hand off
Transfer health questions, urgency, distress, and anything outside the approved scope.
Aimée follows that boundary. With a connected calendar or CRM, it can check available slots, book one, and read the details back. It does not diagnose, recommend treatment, or decide how serious a symptom is.
Can automated scheduling save time without creating a new queue?
Scheduling is not simple. It combines provider rules, limited openings, patient preferences, and a live diary. MGMA's 2026 research describes it as one of the most time-consuming call types for practice teams.[5]
Automation will not repair a broken scheduling process. If appointment types are inconsistent or the diary is unreliable, the caller will feel that mess through any interface.
When the underlying process is clean, the evidence is stronger. A scoping review of 30 peer-reviewed studies found automated self-scheduling produced consistent gains in staff time, patient satisfaction, and appointment attendance.[6] A separate private-practice study reported better slot utilisation and a lower no-show rate after online scheduling was introduced.[7]
The best operational examples are similarly narrow:
- UAMS automated 95% of after-hours cancellation calls, saving more than 800 staff hours a year.[8]
- An NHS England pilot used AI to improve scheduling and reminder logic. It prevented 377 no-shows over six months and enabled 1,910 additional patients to be seen.[9]
- Banner Health reported 70% less manual message triage and faster responses, while keeping staff involved where judgement was needed.[10]
None of these examples proves that every clinic should automate every call. They show that repetitive administrative work can move out of the queue without moving clinical responsibility with it.
Does this replace the front desk?
The strongest pattern is augmentation: after-hours coverage, overflow, voicemail clean-up, booking, and repetitive callbacks.
That is different from saying staffing never changes. A clinic that removes hours of manual phone work may hire differently later. It would be dishonest to promise otherwise. But the immediate operational gain is usually simpler: the team has more time for the patient standing in front of the desk.
There is also a real limit. A human-only answering service may be the better primary option for clinics where calls are routinely emotional, complex, insurance-heavy, or strongly preferred to be human by the patient population. An AI layer can still cover overflow. It should not become the only route in.
For a dimension-by-dimension view, see the honest comparison of an AI receptionist and an answering service.
A better way to evaluate a trial
Do not ask only whether callers “like AI”. Review twenty real calls and score the moments that shape the experience.
If those answers are consistently yes, the technology is doing its job. If one answer is repeatedly no, that is a design problem to fix before routing more calls.
You can also estimate the value currently sitting inside missed calls before deciding how much coverage the clinic needs.
FAQ
Do patients complain when a clinic switches to an AI receptionist?
Some do. The complaints cluster around the same failures covered above: no access to a person, poor understanding, hidden disclosure, or a system operating beyond admin tasks. A disclosed system with fast handoff creates much less friction than one trying to handle everything.
Is it acceptable not to tell patients they are speaking to AI?
It is a bad trust decision even where local rules may allow it. Aimée calls open with a short disclosure so the patient knows what kind of interaction to expect.
Will older patients refuse to use an AI phone system?
Some will prefer a person, and the KLAS research identifies that as a real segment.[1] The response is not to hide the automation. It is to keep a fast human route for anyone who asks.
What happens if a patient gets upset or asks for a human?
The call should move to a person or an agreed callback flow. Distress, health questions, direct transfer requests, and anything outside the clinic's approved knowledge are handoff conditions, not moments for the system to improvise.
Does an AI receptionist make a clinic look impersonal?
It can, if the call becomes a maze. A disclosed system that resolves a routine booking quickly can feel more attentive than voicemail or a long hold. The experience depends on the flow, not the label.
Is there a clinic where this approach is a poor fit?
Yes. Clinics with predominantly complex, emotional, or insurance-heavy calls may need a human answering service as the main route. Automation can still support overflow and after-hours work, but it should not be forced into calls that need judgement.
Sources
- KLAS Research + Luma Health, Patient Perspectives on AI for Healthcare (2025): research report
- Sogolytics, Healthcare AI Adoption and Trust Report (2026): research report
- JMIR, qualitative study on patient attitudes toward AI support tools (2025): study
- JAMA Network Open, international patient survey on AI in healthcare (2025): study
- MGMA, Phones Are Still a Backlog Costing Medical Practices Time (2026): practice research
- Scoping review of automated patient self-scheduling: PMC
- Frontiers in Digital Health, online appointment scheduling study (2025): study
- Luma Health, UAMS case study: case study
- NHS England, AI scheduling pilot (2024): announcement
- Luma Health, Banner Health case study: case study