AI in medical documentation: opportunities and new requirements

The Innovation Sandbox of the Canton of Zurich demonstrates how AI language models (LLMs) and automated speech recognition are now revolutionizing medical documentation processes – while also raising new legal and ethical questions. Below is a summary of the key points.

Efficiency gains through AI

Speech recognition and LLMs are now able to transcribe and structure findings and reports accurately within minutes. Especially in times of staff shortages, this can ease the burden on medical professionals, significantly reduce administrative workload, and minimize potential sources of error. More time can thus be dedicated to direct patient care – improving overall quality of care.

Quality and the need for oversight

Even though AI simplifies processes, the human factor remains indispensable. Without careful validation, errors or factual misinterpretations may arise, which in the worst case could cause direct harm to patients. The rule still applies: AI-generated content must never be adopted uncritically. Oversight by qualified medical staff is essential.

Data protection, professional confidentiality, and cloud use

Every application of AI must comply with strict requirements for data protection and confidentiality. Health data, in particular, is considered highly sensitive personal data. The use of cloud services for storing and processing such data requires detailed contracts and robust technical safeguards – such as end-to-end encryption and strict separation of access rights. Legal certainty decreases when data is stored abroad or with U.S. providers. Professional secrecy requires that confidential information be processed only by authorized individuals; in the age of AI and cloud computing, new organizational and technical solutions are needed.

Regulation and responsibility

It is increasingly unclear whether new AI-based documentation solutions should be classified as support tools or as medical devices – with significant consequences for compliance and liability. A product- and context-specific assessment is already mandatory before AI systems may be integrated into clinical operations. The adoption of best practices such as modular system architectures and model cards for transparency is recommended.

Recommendations from the Innovation Sandbox for AI:

• Early assessment of whether the AI solution qualifies as a medical device and which risk class it falls into.

• Contract-based assurance of all data protection and security requirements when using cloud services.

• Consistent technical and organizational safeguards – particularly end-to-end encryption.

• Careful oversight, validation, and transparent traceability of AI-generated content by medical professionals.

• Avoidance of “shadow use” of untested tools; oversight and clarity in the use of innovative AI solutions enhance security and compliance.

• Regular training and continuous education for staff to ensure the correct use and verification of AI solutions.

• Collaboration with specialized advisory bodies for early legal classification, as well as with authorities and patient organizations for continuous evaluation of solutions.

Praxisbox: m+p develops an application for the automation of patient administration together with sAIpient

The aim was to build an app and a web application that doctors can use to record and transcribe their patient conversations. The application uses AI to analyze complex patient cases, automates billing processes, and enables dictation and the inclusion of notes.

The application for all channels has been launched. Over 250 doctors already use Isaac in their daily work. With various connections to PIS (Practice Information Systems), and more are planned.

Conclusion

Outdated systems in hospitals and surgeries are a real - and often underestimated - risk to patient safety. Digitalization can only lead to better results if it is implemented responsibly, in compliance with data protection regulations and using the latest technology. The use of AI solutions in medical documentation offers opportunities to reduce workload and improve quality, but also poses new, complex challenges. The protection of patients must always remain the top priority. Only those who consider technology, law, and ethics together will be able to exploit the opportunities of digitalization in the healthcare sector safely and sustainably.