AI Agent Orchestration Platform · View 01 of 32 · 1 · Context and scope
In scope
- Defining, versioning and publishing agents, workflows, tools, models and policies
- Executing single-agent and multi-agent workloads, synchronous and asynchronous
- Governing what an agent may use, spend and do, and proving it afterwards
Out of scope
- Model training and fine-tuning — Azure Machine Learning remains the home for that
- Systems of record — the platform acts on SAP, Salesforce and ServiceNow, it does not replace them
- Content authoring and knowledge curation — the platform indexes the existing estate
Assumptions
- Microsoft Entra ID is the single identity provider for people, services and agent identities
- Azure AI Foundry is the default inference surface; other providers are reachable but exceptional
- Purview sensitivity labels already exist on the content estate and can drive routing policy