How Microsoft Foundry fits our AI product direction
Microsoft Foundry, formerly Azure AI Foundry, is our intended AI platform for most future releases. It supplies building blocks; our product work must turn them into useful interfaces, bounded actions, and clear customer workflows. Our current Dashboard prototype does not yet integrate with Foundry.
Choose the task before choosing an agent
A single model call may be enough for a text task. An agent becomes useful when the task needs tools, retrieval, or orchestration. Foundry provides declarative prompt agents and hosting for custom agent code. We intend to choose the simplest architecture that can meet a product's requirements.
Distinguish model knowledge from business context
A model cannot know a customer's private records merely because it is hosted on Azure. Foundry IQ can retrieve from configured knowledge sources and provide citations. The selected sources, permissions, and identity integration determine what the application can access. Each product must document its own configuration.
Treat tool access as product behaviour
Tools and APIs can let an agent request actions in a business system. A useful product needs explicit permissions, error handling, and a clear review point for consequential actions. A platform connector alone does not establish that an integration is shipped in our software.
Check feature readiness before making a claim
Microsoft's new Foundry portal has generally available core experiences, while several advanced capabilities remain in preview. Foundry IQ has mixed API and portal availability. We distinguish platform feature readiness from our product implementation status. These references were reviewed on 10 October 2026; availability can change.
Explore our AI technology approach