Microsoft · AI-500
Performance, context windows, tokens, and cost
AI-500 · Lesson 12 of 16 · 1:38:35
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: task duration and the critical path, parallelism within rate limits, throttling and provisioned throughput, the four context window failures (sliding-window amnesia, summary drift, vector-only recall, entity continuity), token limits, loop controls, and tool call costs, and cost management with Cost Management, quotas, API Management token limits, budgets, and chargebacks. Includes three original practice questions.
Preview the topics
- After two hours of planning, Contoso’s assistant mixes up two travelers on the same family booking and applies the wrong passport details. Which issue is this, and what fixes it?
- One Contoso workflow occasionally runs up a huge token bill, because an agent keeps replanning and calling tools for an hour. What control is missing?
- Three Contoso teams share one model deployment, and finance wants each team charged for its own token use. What should the platform team implement?