An agent can consume Copilot Credits in Work IQ, capacity in Fabric, and tokens or PTUs in Foundry. Examine the meters, a calculated example, and how to separate consumption, attributed cost, and the invoice without counting an operation twice.
The demo convinces, the workspace disagrees. In Microsoft Fabric the ontology is now a real item: a REST API, a definition in files and a child graph created on its own — and it is that child that issues the bill. This article gathers what the item really is, why every refresh is full and freshness became a cost decision, the four capacity meters, how far binding actually reaches, three points where the official documentation contradicts itself — and what has to be in place before switching this on in production.
Federating, mirroring, and ingesting are not synonyms — and in Fabric "mirroring" already means three different mechanisms. When to use Lakehouse Federation, when to mirror a catalog or a database into Microsoft Fabric, when to ingest with Lakeflow Connect, what is actually free, and why source permissions do not cross the OneLake boundary.
Prompt caching solves reuse, but it does not decide when to run, how to buy capacity, or which model should answer. A practical framework for combining Batch, PTU, pay-as-you-go, and Model Router around cost per approved task.
The bottleneck of enterprise AI is not the model, it is governed business context. How Unity Catalog Metric Views, Genie Ontology and Fabric IQ form a single semantic layer over one copy of the data — with an FSI use case, cost model and roadmap.