AI-103: Developing AI Apps and Agents on Azure

Microsoft Foundry Agent Service workflows provide a low-code, graph-based way to orchestrate multiple agents through a deterministic, step-based process: executors (agents or logic steps) are connected by edges that support conditional branching (if/else) and can pass shared state between steps, including an optional downstream action such as invoking a ticketing tool based on an earlier agent's triage result. This satisfies the requirement for deterministic branching with shared state while minimizing development effort, since the branching and state-passing logic does not need to be hand-written. Using threads and runs directly, or coordinating separate agent runs in application code, requires custom-written orchestration logic, and a multi-agent group chat session is a conversational, non-deterministic collaboration pattern rather than a deterministic branching process.