Every new agent usually starts with the same work: another index, another chunking strategy, another set of credentials, and another retrieval pipeline to maintain. In this ESPC webinar, David Lorenzo and I show how Foundry IQ changes that by giving you one reusable, permission-aware knowledge layer that many agents can share, whether they run in Microsoft Foundry Agent Service or in Copilot Studio. We start with the concepts, including what a knowledge base is, how knowledge sources differ from a classic single-shot RAG index, and why agentic retrieval gives better answers on unclear, multi-step questions by planning the query, running sub-queries in parallel, reranking the results, and searching again when needed.

After the theory we build everything live. We create a knowledge base with multiple knowledge sources, tune the retrieval prompt and reasoning effort to balance quality against latency, and secure it with managed identities, Entra ID, and document-level permissions instead of keys. Then we connect two very different agents to the same knowledge base, one pro-code in Foundry and one low-code in Copilot Studio, and compare their grounded, cited answers side by side while each user only sees what they are allowed to see. We close with lessons learned, common configuration mistakes, and cost and latency points to plan for before production. You can watch the full recording on the ESPC Learning Hub.