Track 03 · Advanced
Agentic AI workflows
Design, ship, and reason about agentic systems — tool use, planning, evals, and the guardrails that keep them honest.
Agents are where AI stops answering and starts acting. This advanced track covers how to decompose a goal across planner and worker agents, hand them tools without handing over the keys, evaluate whether they actually worked, and build the guardrails that keep autonomous systems safe in production.
What you'll be able to do
- Decompose a goal across planner and worker agents
- Scope tools and permissions so agents can't overreach
- Build evals that tell you whether an agent actually worked
- Design guardrails and human checkpoints for autonomy
Inside the track
- 01
Delegation basics
The planner decomposes and routes; specialist workers execute. Why that split keeps context narrow.
- 02
Tool use
Giving agents real capabilities — and scoping them so a mistake stays contained.
- 03
Planning & orchestration
Coordinating multi-step work across several agents without it descending into chaos.
- 04
Evals
Measuring whether an agent succeeded, not just whether it ran to completion.
- 05
Guardrails
Bounding fan-out, cost, and permissions; keeping a human in the loop where it counts.
Start your streak on Agentic AI workflows.
Take a free starter lesson on the web, then carry your streak straight into the mobile app and keep the daily reps going.