A structured, transparent way for Swiss hospitals, practices and other health providers to decide whether an AI use case should be deployed, held or rejected.
Start the assessmentSee a worked exampleAdoption models such as TAM, UTAUT and TOE explain whether a technology is accepted in general, but they give no criteria for a specific AI use case in a specific organisation. Generic prioritisation tools compress everything into one score and assume a tool performs the same everywhere. That does not hold across 26 cantonal systems and three very different provider types.
The framework is built for hospitals (NOGA 861), medical and dental practices (NOGA 862) and other health providers (NOGA 869).
The sun radiates the same energy whatever the weather, but what reaches the ground depends on the atmosphere. An AI application works the same way. Its potential impact is a property of the tool; what it actually delivers depends on the organisation it is deployed into.
So every use case is assessed on two independent dimensions: feasibility (do today’s conditions allow it?) and stakeholder impact (what does it do to patients, staff and the organisation?).
Each domain asks one guiding question and has three sub-domains. Point at a segment to highlight a domain and its three sub-domains; the guiding question appears in the centre. All eight are described in full below the wheel. Each domain has a target: the level the AI application would have to reach in that area.
For each sub-domain the team also records whether an AI project that is already running or planned in parallel makes the condition easier or harder to meet. A documentation AI already in use may have set data governance standards that a new imaging AI can reuse; or it may have used up this year’s digital budget. Where no other project touches a sub-domain, the influence is zero.
Answer Yes or No for the conditions of the eight domains. Questions that do not apply to the use case are left out automatically.
For each affected group, record benefits and disadvantages separately and whether serious harm is possible. Nothing is averaged into one number.
The result appears immediately, with the reason and the open points. A weakness in one place is never balanced out by strengths elsewhere.
Patients, clinical staff and the organisation are always assessed. Payers, authorities and other groups are added when the use case is likely to affect them.
For each group, the clinical, operational and economic effects are recorded, each as a benefit and as a disadvantage. Serious harm to a group can never be outweighed by benefits elsewhere.
The conditions are met, the overall impact is positive and no serious harm remains open. For clinical, regulated, costly or hard-to-reverse use cases, named experts review the listed points before go-live.
One or more areas are not ready, or open points remain. The result names what is missing, so the team can improve the conditions and set a point for reassessment.
The overall impact is negative, or a serious harm cannot be prevented, regardless of how strong the feasibility conditions are.
How the framework reads a real-looking situation. A hospital evaluates an AI tool that analyses chest X-rays to support radiologists. It already runs one AI project: a clinical documentation tool, live for six months.
Radiologists are cautious, but have seen the documentation AI work reliably.
A mature image archive with ten years of imaging, but only partial EHR interoperability. The governance standards of the documentation project transfer directly.
An AI oversight committee already exists and the medical device classification is underway.
The documentation AI used up a large share of this year’s digital budget.
Positive for staff and the organisation, positive for patients overall, but with a possible serious harm: lower accuracy for an underrepresented patient group in the training data.
Complete an assessment alone or as a team. Everything runs in your browser.
The empty tool for your own AI use case: answers, result and dashboard in one place.
A finished assessment (AI-assisted chest X-ray triage) to see how the tool and its result look.
Paste the export codes of your team members to see a preliminary group result and where answers differ.
A consolidation of six fictional team members, with group graphs and the dashboard of each member.
The framework and tools were developed by Adelia Safina as part of the Master Thesis 2026/27 (MSc Business Information Systems, FHNW School of Business).
The result supports a decision. It does not replace the judgement of the responsible clinical, legal and data protection bodies.