Most Fabric estates have already done the hard part. The data is landed, modelled and governed. The value is still stuck in the last mile, where someone has to open a report, notice the thing, interpret it, and go do something about it.
Agents collapse that last mile.
This is not a chatbot demo. Microsoft has been rebuilding Fabric as the grounding layer for agentic AI. Data agents query your lakehouses, warehouses, semantic models and Eventhouses in natural language. Fabric IQ adds a shared context layer of semantic models, ontologies and graph, so an agent understands what a customer actually is instead of just seeing tables. Operations Agents watch real-time signals continuously and take approved action. That same knowledge then reaches into Copilot Studio, Microsoft Foundry, Teams and M365 Copilot. Governed once, reused everywhere.
Kevin Harmer, ArchitectNow's Chief AI and Data Officer, builds the whole picture in sixty minutes and spends half of it in the product rather than on slides. Four live demos on real data, plus an honest read on what is production ready today and what is still preview.
You'll walk away with:
- A clear mental model of the four Fabric agent building blocks (Data Agents, Fabric IQ, Operations Agents, and outbound integration) and when to reach for each.
- A working pattern for grounding an agent in your own semantic layer, so it answers in your business's language rather than your schema's.
- A repeatable action loop: monitor a real-time condition, recommend a response, and execute it with the right approvals and identity.
- A reuse strategy for exposing Fabric knowledge to Copilot Studio and Microsoft Foundry without duplicating logic or loosening governance.
- An honest preview versus GA read on what is production ready today and what to pilot, plus a 30/60/90 adoption path.
Built for data and analytics practitioners: data engineers, analytics engineers, BI and Power BI developers, data architects, Fabric administrators, and AI/ML engineers.
Also built for the technical leaders asking them "so when do we get agents on this?" and wanting a straight answer.
Helpful background: working familiarity with Fabric workspaces and OneLake. No prior agent-building experience assumed, but we will not spend the hour explaining what a lakehouse is.
Kevin Harmer, Chief AI & Data Officer at ArchitectNow. Hosted by Craig Brandt.
Part of Getting AI Into Production on Microsoft, ArchitectNow's webinar series running through December.
Next session: Stop Chatting, Start Automating: The Business Workflows AI Should Already Be Running. Tuesday, September 29 at the Microsoft Innovation Hub.