The interoperability of health information systems today determines the relevance of any solution intended for healthcare professionals. Tools that do not integrate with the DMP, the secure messaging MS Santé, or the Ségur du numérique reference frameworks create disruptions in the care pathway, regardless of their level of technological innovation.
Ségur Interoperability and DRIMbox: the technical foundation that tools must comply with
Wave 2 of the Ségur du numérique changes the game for medical imaging. The goal is no longer just to share reports between establishments, but to make the medical images themselves accessible via the DRIM-M network and DRIMbox. For a radiologist or a surgeon, this means direct access to a colleague’s images without burning CDs or manual transfers.
This level of technical requirement filters market solutions. A business software that does not comply with the Ségur reference frameworks will no longer be eligible for public funding associated with tool updates. This funding is temporary, which pushes practices and establishments to quickly make decisions about their equipment choices.
Professionals looking to map the stakeholders and sectors involved in this transformation can consult the website www of La Santé de Demain, which organizes information by specialty and type of need.
Generative AI in clinical practice: real time savings or additional verification burden
Generative AI is entering healthcare facilities through the back door. A national survey conducted among more than a thousand hospital staff in 2025 reveals that some caregivers were already using personal generative AI tools, in the absence of dedicated institutional training. This observation deserves attention: adoption precedes regulation.
The promised time savings in writing reports or synthesizing patient files face operational reality. Each output from the AI requires careful proofreading. For a general practitioner handling several dozen consultations per day, this verification task can consume a significant portion of the time saved upfront.
The four HAS reflexes for using generative AI
Since October 2025, the HAS has published practical guidelines summarized in four verbs:
- Learn: understand how the tool works and its limitations before integrating it into a clinical workflow
- Verify: systematically check the generated outputs, especially factual medical data
- Assess: evaluate the benefit-risk ratio of usage in each care context
- Communicate: inform the patient and colleagues when AI has contributed to the production of a document
These recommendations are not legally binding, but they establish a framework for professional responsibility. A practitioner using an AI tool without applying these reflexes exposes themselves to difficulties in case of disputes regarding the quality of an act.
Medical voice assistants and scribes: reducing administrative time without degrading the patient relationship
Digital scribes and medical voice assistants represent the most concrete category of tools to reduce administrative burden. The principle: the tool transcribes and structures the consultation in real-time, while the practitioner remains focused on the patient.
Their deployment raises two technical questions that commercial presentations often overlook. The first concerns data hosting. A voice assistant capturing the entirety of a medical exchange produces health data in the sense of the GDPR and the HDS framework (health data hosting). The provider must be certified, and the practitioner must be able to justify the compliance of their choice.
The second concerns the quality of transcription in specialized medical vocabulary. A dermatologist and a psychiatrist do not have the same lexical requirements. Generalist solutions produce transcriptions usable in general medicine, but their error rate increases as one enters a specialized field with its own terminology.
Cybersecurity of practices: an organizational requirement beyond software
The proliferation of digital tools in practices and establishments creates an expanded attack surface. Cybersecurity in health is no longer a purely IT issue; it is an organizational constraint that affects every professional on a daily basis.
The NIS 2 directive, now transposed, extends security obligations to a broader range of actors in the healthcare system. Concretely, a group practice or a multi-professional health center must formalize incident management, backup, and access control procedures, even without a dedicated IT service.
Points of vigilance for local care structures
- Ensure that each business software has strong authentication (not just a shared password among partners)
- Ensure that backups are outsourced to a certified HDS host, with at least annual restoration testing
- Train all staff, including medical secretaries, on targeted phishing techniques specifically aimed at the health sector
The cost of a cyberattack on a practice far exceeds any potential ransom. Business interruption, loss of patient data, and damage to professional reputation are harms that we are increasingly witnessing in the sector.
Innovative solutions that support healthcare professionals daily are not just a technological race. The most efficient tool remains useless without interoperability, regulatory compliance, and team training. Equipment decisions should be guided by these three criteria before any consideration of functionality.
