Testing AI systems with the organization's own data can be a way of assessing the suitability of the AI system in this phase. This allows for technical testing related to data flow and infrastructure compatibility, as well as evaluating the AI model’s performance using the organization's own data.
It may also be relevant to consider a retrospective validation process at this stage. If the suppliers permit this, it is important to ensure that all vendors are treated equally in accordance with procurement regulations [118]. See Phase 5 for legal grounds for data processing and exceptions to confidentiality requirements regarding the use of data from the organization’s medical records. If validation is desired but not feasible before the contract is signed, the organization may consider including a contractual provision allowing for termination if the AI system does not perform as expected when validated with the organization's own data.
[118] As also described in phase 2, a situation where a solution has been fully tested (including integrations) from a single supplier before initiating a competition phase could create challenges in terms of conducting a procurement that fulfils the requirement for equal treatment in the procurement regulations. In such situations, it is necessary to equalise any advantages that a supplier may have gained, which in such cases often means that the procurement process can take quite a long time, because the other suppliers must also have the same knowledge base and should have the opportunity to showcase their solutions in a similar way before a contract is awarded.