Microsoft 365 Copilot now draws business context from Fabric IQ by default

Arun Ulag, Microsoft's executive vice president for Azure Data, wrote in his announcement post that Fabric IQ combines "unified data from OneLake, trusted metrics from Power BI semantic models, and operational context from ontologies and real-time intelligence" ^[7]. It is one of four parts of what Microsoft calls Microsoft IQ. Work IQ covers email, Teams and SharePoint, Foundry IQ covers documents and manuals, and Web IQ covers the public internet ^[6]. Of the four, only Fabric IQ holds metric definitions ^[20].

I think the design is right. Each metric is defined once in Fabric and read from there by Copilot, by outside agents over MCP and, soon, by apps Power BI builds from the same definitions ^[2][8][3]. One definition in one place is easier to audit and fix than a copy maintained per agent.

"Agents are a very, very strange animal, and I always like to say it's like Drew Barrymore from 50 First Dates. Every time they open their eyes, they forget everything that happened before," Amir Netz, Fabric's CTO, said in a press briefing after the keynote ^[4]. As keynote metaphors go, it is an accurate description of a stateless process. Ulag said coding agents work because they have "the code, the repos, the change history, the specs, the tests," and argued that most enterprises have nothing comparable outside coding ^[5].

Ulag's list ends with tests. Yitzhak Kesselman, the corporate vice president who runs Fabric IQ, described evals as something advanced customers build for themselves. "Customers that are more advanced on their journey have their own evals for their agent," he told The New Stack ^[10]. What those evals show sometimes isn't what customers expected, he said ^[11]. He met with more than 320 companies last year ^[12].

The account from Barcelona does not say whether an administrator can switch the Copilot default off, or which identity an outside agent's MCP query runs under ^[2][8]. For tables mirrored from Snowflake, keeping the source's access roles depends on mirrored security roles, which are still weeks from public preview ^[16]. "We're not just solely replicating security rules. We're actually actively enforcing and preserving those roles," said Shireen Bahadur, who demonstrated the feature in the keynote ^[17].

Microsoft does not charge for shortcuts and mirroring, the features that connect or replicate external sources into OneLake ^[13]. At the 300% annual growth Netz cited, OneLake now manages four times the storage it did a year earlier ^[19]. The dbt Fusion engine is due in Fabric in the next couple of weeks ^[18].