Paolo PedrazziLearning to direct AI agents
Courses and workshops on agentic AI: how to give an agent a mandate, how to verify what it produces, how to put it into production without getting hurt.
I don’t teach prompt writing. I teach the craft that remains when code costs almost nothing: choosing what to build, giving instructions that hold up, and telling a good result from a plausible one.
Open courses
Bringing agents into a team
Custom programs for companies that want to stop experimenting and start putting agents to work on real processes. In person or remote.
What we cover
- Where to start: which processes can really be delegated
- How to write a mandate an agent can actually execute
- Verifying the work: tests, review, acceptance criteria
- The real costs, limits and risks of a fleet of agents in production
How it works
- A program tailored to your processes, not an off-the-shelf course
- We work on your cases, with your data
- Every session leaves behind something that runs, not just notes
- Support in the weeks after, when the real problems show up
What you take home
The mandate
A method for assigning work to an agent: goal, constraints, acceptance criteria. It’s what makes the difference between a toy and a collaborator.
Verification
A way to verify what the agent delivers without rereading it line by line: tests, automated checks, sampling. Delegating without verification isn’t delegating.
The first piece
At least one real piece of your own work that runs on its own by the end of the course. Not an exercise: something that’s already in production the following Monday.
One hour to figure out where to start
If you’re not sure which course you need, book a session: we’ll look at your situation and decide together whether a program makes sense, a project, or nothing at all.