AI · 20 November 2026
Microsoft Foundry IQ Adds Knowledge Graphs for AI Agent Grounding
Microsoft has added knowledge graphs, a specialised retrieval model and automatic infrastructure provisioning to Foundry IQ, helping developers build AI agents that reason over enterprise data. The knowledge graphs were announced at Microsoft Ignite.
What happened
Microsoft has added knowledge graphs, a specialised retrieval model and automatic infrastructure provisioning to Foundry IQ, its knowledge retrieval service. The update is designed to make it easier for developers to build AI agents that can access, understand and reason over enterprise data.
The knowledge graphs were announced at the Microsoft Ignite conference. In a blog post, as reported by InfoWorld, Microsoft said they help agents connect related entities and facts across documents and organise enterprise content into topics. This allows agents to answer questions that draw on information from multiple sources.
To map relationships across documents, developers can add a graph configuration containing their ontology to an existing Foundry IQ knowledge source. Microsoft positions the change as a way to ground agents in accurate, relevant and up-to-date business information.
Why it matters
Grounding is the constraint that often decides whether an AI agent can be trusted with real work. As organisations automate workflows, an agent that retrieves a single relevant passage is far less useful than one that can follow the links between a contract, a policy, a product record and a customer history. Knowledge graphs move retrieval from finding documents towards understanding how the information fits together.
For leaders in digital transformation, the change also shifts where effort goes. Defining an ontology, meaning the entities, relationships and categories that describe the business, becomes a design task that spans domain experts and developers. Teams that treat it as an afterthought are likely to find that agents answer confidently but connect the wrong facts.
The Renascence take
Most coverage of this announcement will focus on the features. The more useful question is who defines the ontology, and whether they understand the customer journeys the agent is meant to serve.
A knowledge graph is only as good as the model of the business behind it, and that model is usually built by people who have never sat with a frustrated customer. Retrieval quality will matter less than whether the relationships reflect how customers actually ask questions and how staff actually resolve them. Operators should start by mapping the top journeys and the decisions inside them, then build the ontology around those. Only after that should they test agents against real service scenarios, not benchmark answers.
Sources
This briefing was written by our Newsdesk, synthesising reporting from the outlets below. Follow the links for the original coverage.
FAQ
Questions we get on this topic
More in AI
Stay ahead of CX
Get the signal, not the noise.
The stories shaping customer experience — plus the Journal and Experience Loom — in your inbox.


