ai risk classes

AI risk classes: quick classification for teams

Getting the classification right matters for the roadmap. It decides effort, evidence and time to production.

Minimal risk: few regulatory obligations; the focus is on transparency and quality.
Limited risk: clear information duties toward users.
High risk: strict requirements for data, documentation, monitoring and governance.
Unacceptable risk: prohibited; rule it out early in scoping.

Trust signal: We link the regulatory classification directly to architecture decisions and the delivery plan.

What an enquiry sets off

The reply comes from the engineer who would do the work, not from a sales layer that passes it on.
A few sentences are enough for a first read. System access, architecture diagrams and an NDA come later, and only if this goes anywhere.
No is a possible outcome. If the problem is solvable in-house, or there is no slot free right now, the reply says so.