Who this is for
- Principal dentists
- Practice managers and DPOs
- Clinical governance leads
AI often arrives without a formal decision: imaging detection features, note drafting, transcription, marketing tools or Copilot. The governance problem starts when someone asks for the evidence file and there is none. The paid dental governance service page covers engagement scope and fees; this guide lists the evidence categories to hold on file.
Where AI appears in dentistry
Radiograph interpretation support. Intraoral scan analysis. Clinical photography. Note drafting and clinical transcription. Patient communication and recall automation. Marketing copy and image tools. ChatGPT and Copilot on the admin desk. AI features switched on inside supplier platforms you already pay for.
The evidence file
Tool level
- AI tool and use case inventory, covering clinical, admin and marketing workflows
- Approved, conditional and prohibited use position
- Named owner for AI governance in the practice or group
Data protection
- Patient-data exposure note: which tools touch radiographs, photographs, notes, correspondence, identifiers
- DPIA screening record, and a DPIA where it is likely required or strongly indicated
- Vendor data processing agreements
- Hosting and data residency confirmation
- Retention and deletion terms
- Sub-processor list
- Privacy notice wording that mentions AI where AI is visibly in use
Clinical
- Documented clinician review step before AI output enters the record
- Incident route covering AI-generated record errors and image-processing concerns
External
- Staff guidance, and evidence it was communicated
- Indemnity and insurer disclosure-readiness indicators
The gaps we see most often in dental practices
AI imaging supplier evidence held by one clinician rather than the practice. ChatGPT and Copilot used with no written patient-data boundary. Transcription running with no DPIA screening. Patient transparency wording that changes depending on which surgery the patient walks into. Marketing AI with no separation between marketing data and clinical data. No approved or prohibited use position at all.
These are evidence gaps, not conclusions about breach. They are also the gaps that take longest to close under pressure, which is why they are worth closing before the request arrives.
Next step
The Clinical AI Exposure Diagnostic™ builds this file for a practice or group in four working days from the start of delivery, with every finding converted into a dated action and a named owner. Multi-site groups are scoped at intake.
Advisory governance support only. Not legal advice, DPIA sign-off, CQC approval, ICO approval, insurer coverage advice, MDO indemnity advice or clinical safety case sign-off. Final decisions remain with the organisation's accountable officers and advisers.