Private AI for medical and dental clinics
The protocols exist, they are up to date, and nobody finds them in time. AI running inside the clinic puts them one message away, with no health data leaving your network.
In a clinic the critical information is written down (protocols, consent forms, memos, equipment instructions) but scattered across folders, email and old versions. Private AI indexes it and answers citing the document it came from. It runs on a server in the clinic, so health data, a special category under GDPR art. 9, is never transferred to an outside provider.
The real pains
Front desk and treatment room, not IT.
The protocol exists, but which version applies?
Three copies of the same document in three folders, and the one in force is whichever someone remembers.
Reception answers the same thing all day
How to prepare for a test, what insurance covers, what to bring to a first visit: same questions, manual answer every time.
High staff turnover
Every new hire repeats the same questions for weeks, always to the busiest person on shift.
The team already uses AI on its own
Someone pastes a report into a website to summarise it. Without an internal tool this happens anyway, just outside your control.
What AI does about those four
One answer, with the source in view
It answers from the protocol you indexed and cites the document it came from, with its date when the document carries one, instead of relying on who remembers which copy is the good one.
Takes load off reception
Repeat administrative questions get answered in seconds, from your own documents.
Supports new staff
New joiners ask the system instead of interrupting, and get the clinic's actual procedure.
Gives an internal alternative to ChatGPT on a phone
Once there is a tool of your own that works, there is no reason left to paste patient information into someone else's website.
Health data plays in another league
GDPR treats health data as a special category (art. 9): the bar for legal basis, security measures and impact assessment is higher than for any other data. Every external tool that patient information reaches is one more processor to contract, document and audit.
With private AI that problem isn't managed. It's avoided. The model and the search run on a server in the clinic, with no route to the internet for the content, and the system stays up even if the connection drops, which on a fully booked day is not a minor detail.
What it does not do
It does not diagnose, does not interpret test results and does not suggest treatment: that would be a medical purpose and would turn the software into a medical device subject to CE marking. It does not replace your practice management software or the patient record either. It is a documentation and administrative tool: it finds, summarises and drafts from what is already written in the clinic, always under a professional's review. And today it does not compartment by user: access is protected by a login, but whoever gets in can query everything that has been indexed. If each profile must see only its own part, the scope of what gets indexed has to be narrowed.
Before you ask
Does the AI diagnose or suggest treatment?
No, deliberately. Software with a medical purpose would be a medical device and would need CE marking under Regulation (EU) 2017/745. This is a documentation and administrative tool: it finds and drafts from your own protocols and documents. Clinical judgement always stays with the professional.
Does it touch patient records?
Only if you decide so, and it isn't needed to start. Most clinics begin by indexing protocols, consent forms, internal memos and supplier documentation, with zero identifiable health data. Widening the scope is a later decision.
Why isn't a cloud tool good enough?
Because health data is a special category under GDPR art. 9: the bar for justification, security and impact assessment is higher than for ordinary data. With the server inside the clinic there is no third-party transfer to document and no international transfer to assess.
Can front-desk staff use it?
Yes, and it is usually the first use case with a clear return: questions about test preparation, coverage, paperwork to bring or cancellation policy get answered in seconds, from whichever protocol you have indexed.
Let's start without patient data
A 30-minute call to scope a pilot using protocols and internal documentation only.