Prioritising an urgent case is not triage: the line the AI Act draws

When a dental clinic considers letting an AI answer its phone, a question comes up early and it is a very fair one: “what if someone calls in pain? Is a machine going to decide how urgent it is?”.

The short answer is no, and it is not just a product decision: it is a line the EU AI Act draws clearly.

General guidance, not legal or clinical advice.

What the law considers high risk

The regulation classifies AI systems by risk level, and places in the high risk category systems intended to classify emergency calls or to establish priorities in dispatching first-response services, including patient triage systems in emergency healthcare.

It makes complete sense: deciding who is seen first based on the severity of symptoms is a decision with real health consequences. A system like that carries very demanding obligations on documentation, oversight and quality.

And what a scheduling assistant does

Now, that is not what happens when someone messages their dental clinic saying a tooth hurts. There is no clinical assessment there: there is schedule management.

The difference lies in who decides what:

• Clinical triage would mean assessing symptoms, estimating severity and setting a care priority based on it. That is medical judgement and cannot come out of an automated conversation.
• Prioritising an appointment means recognising that someone reports urgent discomfort, offering them the first available slot and alerting the clinic team so a professional sees it. There is no assessment of severity and no clinical decision: there is a schedule being reordered and a person the matter is passed to.

Put another way: the assistant does not decide how serious the case is. It recognises that the patient is asking to be seen soon, acts on the calendar and puts a professional in the loop.

Why it is worth being clear about this

For two reasons, beyond compliance.

The first is patient safety. An automated reply that hints at an assessment — “that doesn't sound serious”, “it can wait” — is exactly what must not happen. The potential harm is obvious.

The second is trust. A clinic that can explain precisely where the line is conveys far more rigour than one promising that “the AI handles everything”. In healthcare, knowing where a tool stops is part of its quality.

How it works in ClaudIA Dental

ClaudIA Dental never assesses symptoms. When a conversation comes in as an emergency, it does three concrete things: offers the first available slot, alerts the clinic team and logs the conversation so the professional has the context. The decision about the case belongs to the dentist, always.

It is the same logic we apply to any enquiry that steps outside the schedule: when human judgement is needed, the conversation goes to a person. You can also see how urgent cases are organised day to day and what ClaudIA does in a dental clinic.

An AI designed well for a healthcare setting is not the one that decides the most. It is the one that knows exactly which decisions are not its to make.