Knowledge graphs deliver the real-time context enterprises need to make AI explainable

At Intuit, the big AI story isn’t another model launch. It’s the graph underneath the model. Chad Cloes, a staff software engineer at the company, said the security team has built its data platform around knowledge graph technology so scattered systems can be tied together in a way that makes sense to both humans and AI agents.

The practical payoff is speed. Cloes said security analysis once meant logging into seven different systems, using three different credentials and waiting days to understand how things were connected. Now the team is aiming to drive mean time to remediate down by turning those relationships into something queryable. The result, he said, is a shift from days to seconds.

That graph layer has also become part of Intuit’s AI stack. The team has built a GraphQL API on top of its graph platform to support Model Context Protocol servers, and developers can query that setup in natural language. The point is not that the database itself is magic. Cloes was blunt that the specific winner on the backend does not matter much if the data is pulled into a graph, made relevant, and used with real context.

And that is the bigger message here. As language models become more interchangeable, the differentiator moves to the data layer: how an enterprise makes its own information usable, explainable and current. Cloes said the graph gives that contextualization as a byproduct, then hands that context to the LLM — whether that LLM is ChatGPT, Amazon Bedrock, or something else later.