Five sources per agent, answers capped at 25 rows by 25 columns, documented support for English only, and read-only queries: the Fabric data agent has an entry contract no prompt renegotiates. This article gathers the six source categories, the output limits, the asking user's identity, and the two retirements of 26 August — and shows what has to be in place before the pilot becomes a commitment to the business.
When a RAG agent answers wrong, the model is rarely at fault: the right passage never reached the context. This article splits retrieval into three budgets that fail for different reasons — recall, precision, and passage context — and shows what fixes each one: structure-aware chunking, hybrid search with rank fusion, second-stage reranking, and retrieval measured separately from the answer.
A2A connects independent agents; MCP connects tools and resources. Understand tasks, identity, and trust across delegation, then follow a complete Python example to deploy and call an A2A agent in Microsoft Foundry, with RBAC and explicit preview limitations.
Guardrails contain at runtime; evaluation measures over time — and neither does the other's job. Content Safety, Prompt Shields, evaluators that see the tool call and not just the answer, red teaming with PyRIT, and the maturity signal that separates a demo from a system: a deployment blocked by a number.
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.
AI adoption stopped being a technology problem and became an operations and accountability problem. An executive guide to governance, the real dangers of AI, and the shift from application developer to AI engineer.
The memory types of an agent (short/long-term, episodic, semantic, procedural) and how to implement the retrieve/write-back loop in the Microsoft Agent Framework.