Business scenarios
Digitising and structuring medical documentation
Knowledge and documents
TWIN:DESKProblem: Medical documentation exists as scans and printouts
How it works: OCR reads data from scans into a structured form available to authorised staff
What you gain: The patient’s history from scans becomes searchable - for those entitled to it
Assistant for current medical procedures
Knowledge and documents
TWIN:DESK + CDFNEW · SAIE PANTERAProblem: Medical standards and procedures change, and staff use outdated versions
How it works: A central RAG knowledge base that cites the procedure in force; CogOps (CDF-F6) measures how current the knowledge is. Agentic RAG asks again in another round when something is missing, and TWIN:INDEX searches long documents by chapter
What you gain: Staff get the current version of a procedure with a citation, and management knows which procedures need review
Medical documentation by voice
Process automation
TWIN:DESKNEW · SAIE PANTERAProblem: Doctors spend time writing documentation instead of with patients
How it works: Voice chat (STT/TTS), the AI Powered Notes editor and structured answers that put the content into the documentation template. The recording becomes text with labelled speakers, and findings turn straight into tasks with a deadline
What you gain: The doctor dictates and the system arranges it into the documentation template - to check and sign, not to rewrite
Safe AI work with patient data
Security and sovereignty
PROXY:AINEW · SAIE PANTERAProblem: Patient data must not be processed outside the facility’s infrastructure
How it works: A predefined Healthcare sector profile; identifying data detected and masked before every query; an “EU only” switch. Offline Model Registry matches the model to the hardware and starts it from a Harbor package in a network without the internet
What you gain: Staff use AI, and the DPO has a technical guarantee that a patient’s identifying data will not leave the facility
AI with a mandatory human decision
Compliance and oversight
PROXY:AI + CDFProblem: No decision concerning a patient may be made automatically
How it works: Approval gates and a HITL queue for critical actions; an explicit level of agent autonomy and the Human Competence Gate - a human makes and signs the decision
What you gain: AI prepares, a human decides - and the system makes sure the approval is informed and documented
High-risk AI compliance in healthcare
Compliance and oversight
PROXY:AI + CDFProblem: A clinical decision support system is a high-risk system under the AI Act
How it works: ASIA/FRIA report (CDF-F1.5), conformity assessment before putting into service (F5); FRIA/ASIA export and an audit evidence package
What you gain: The facility deploys clinical decision support along a path it can defend before the regulator - with documentation from day one
Questions decision-makers ask
Who is responsible for AI-supported decisions?
The doctor who makes the decision, and the facility as the deployer of the system. AI in SAIE prepares summaries and draft documents with a source citation, and a human always approves the decision
Can the system run entirely within the facility?
Yes. TWIN:DESK and PROXY:AI run on-premises, in the facility’s infrastructure, with local models
How do you divide responsibilities between the director, the DPO, the CISO and the provider?
The director is responsible for the deployment decision and the policy, the DPO (data protection officer) for the legal bases for data processing, the CISO for security and compliance with the KSC Act, and the provider for system documentation and support. The CDF methodology includes templates for these roles and registers
Can a doctor use an AI chat with patient data?
Not in a public AI chat: health data is a special category of personal data (Article 9 GDPR), is covered by medical confidentiality, and the facility does not control where it goes. Yes, if the AI chat for doctors runs in the facility’s infrastructure: TWIN:DESK and PROXY:AI from the SAIE ecosystem run on-premises with local models, PROXY:AI with the Healthcare sector profile masks a patient’s identifying data before every query, and a human always approves the decision
All questions