Project 1000 – Agentic AI: Testing requirements for LLM-based agentic artificial intelligence (PRAKI)

Summary

The tender concerns a study and the development of testing criteria, testing catalogues and best practices for LLM-based agentic AI systems, including multi-agent systems and communication and authentication between agents. The work covers at least three use cases, risk analysis, selection and detailed examination of two use cases, testing procedures, stakeholder trials, a generalised testing catalogue, publications in human-readable and OSCAL formats, a final study targeted at approximately 90 pages, and a virtual final presentation. The project is planned for a maximum of 18 calendar months. Offers must include the required company, reference, personnel, technical equipment, cost and milestone-plan information. The award is based on the best price-performance ratio using qualitative award criteria; each B-criterion requires at least 1 evaluation point.