Microsoft · AI-500
Orchestration, human-in-the-loop, caching, and scaling
AI-500 · Lesson 9 of 16 · 1:08:25
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the AI-500 exam (Microsoft Certified: Multi-Agent AI Solutions Expert). In this lesson: hub-and-spoke (orchestrator-subagent), sequential, parallel, and peer-to-peer orchestration, Agent Framework's built-in workflows, durable human-in-the-loop approvals with overrides, compensating actions, and timeout rules, prompt, semantic, and response caching, and scaling with spawn caps, concurrency sized to quota, quorum policies, checkpoints, and the Azure OpenAI Batch API. Includes three original practice questions.
Preview the topics
- Contoso’s refund agent waits for a manager’s approval by keeping its process running, and every deployment restart loses the pending refunds. What should the design use?
- Contoso’s agents share a 3,000-token system prompt, but each request starts with the traveler’s message, and prompt caching never kicks in. What should the team change?
- Each week, Contoso re-scores 50,000 hotel reviews for a report due two days later. Which approach fits?