What buyers usually need to see
Procurement teams are rarely asking for abstract AI principles. They need artefacts they can file, review, and compare. That usually means an AI system inventory, a data and vendor map, an EU AI Act role and risk view, oversight notes, security alignment, and a repeatable way to answer questionnaires.
The evidence pack to prepare first
Start with the evidence that unblocks commercial conversations: system name, intended purpose, owner, user group, data categories, vendor or model provider, hosting location, risk notes, human oversight, monitoring route, and the latest review date.
How to use this checklist
Use the checklist before a customer asks for it. The best moment to find gaps is before a deal, investor review, or board meeting turns AI governance into an urgent request.
Practical evidence checklist
- List all product AI features, internal AI tools, models, APIs, and vendor platforms.
- Assign a business owner and technical or operational owner to each AI use case.
- Map whether the organisation is acting as provider, deployer, distributor, importer, or AI user.
- Record whether the use case may fall into prohibited, high-risk, limited-risk, GPAI, or lower-risk scenarios.
- Map personal, confidential, customer, and sensitive data used by each AI system.
- Keep vendor, model, hosting, API, and subprocessor evidence in one place.
- Document oversight, monitoring, escalation, and incident routes.
- Align AI evidence with GDPR, ISO 27001, security review, and ISO 42001 readiness.
- Prepare a short procurement response pack that commercial teams can reuse.
FAQ
Is this the same as being EU AI Act compliant?
No. Procurement readiness is not a legal certification. It is a practical evidence position that helps teams answer buyer, investor, board, and security-review questions. Legal advice may still be needed for classification and obligations.
Who should own the checklist?
Ownership is usually shared between product, technology, legal, security, data protection, and commercial teams. One person should own the evidence pack, but the inputs come from several functions.
How often should the evidence be reviewed?
Review it whenever an AI system, data flow, vendor, model, use case, or customer commitment changes. A quarterly review rhythm is a practical starting point for many teams.