AI-102: Microsoft Agent Framework

Hi,

Short answer: study against the official AI-102 skills outline, not against whatever the newest lab happens to use this month.

Right now the live exam is still aligned to things like:

- Azure AI Foundry Agent Service

- Semantic Kernel

- AutoGen

Under the “Implement an agentic solution” section (create agents, more complex workflows, multi-agent, testing, deploying, etc.).

Microsoft Learn content (with Microsoft Agent Framework / the agent_framework library) usually moves faster than the exam. Learn gets updated first, the question bank lags behind by a few months.

So in practice:

Yes, you should understand the agentic concepts the new labs teach (agents, tools, workflows, orchestration, evals).

No, the exam has not suddenly become “Microsoft Agent Framework only”.

Officially it’s still framed as Azure AI Agent Service + Semantic Kernel + AutoGen.

If I were studying right now, I’d do this:

Use the new Microsoft Agent Framework labs to learn the modern way of building agents.

Map everything you learn back to the “skills measured” for AI-102: Agent Service, SK, AutoGen, multi-agent patterns, evaluation, deployment.

Don’t waste time memorizing every API call in agent_framework; focus on architecture, scenarios and design decisions, because that’s what exam questions tend to test.

So: anchor on the official study guide, use the new Agent Framework content as a nicer, more future-proof on-ramp to the same underlying exam skills.