Project 1 · Models and context
Teach AI the job
A model answers a company’s customers badly and makes things up. You give it a role, the company’s context and the rules it owes its customers, then run the evals until every ticket passes.
Pick your seats this week. Spend one day at your offices building with AI. Leave with every coworker signed off by a peer.

10:42 · Project 1
Lea raises a hand.Tom takes it and sits beside her.Her tests pass.
The method of 42, run on more than 50 campuses, including
Paris
Tokyo
Berlin
Madrid
Lisboa
Abu Dhabi

Project 1 · Models and context
A model answers a company’s customers badly and makes things up. You give it a role, the company’s context and the rules it owes its customers, then run the evals until every ticket passes.
Project 2 · Tools and MCP
Connect a model to company tools through MCP and plan an offsite with them. You decide which tools it may use alone, which need your approval and which are off limits, then approve or refuse each action live.
Project 3 · Building tools
Build a real MCP server for a fact-checking desk, with tools over Wikipedia, earthquake records, past weather and exchange rates. You write what each tool does so the model uses it right, then check articles nobody has seen yet.
Group project · Working as a team
The whole group ships one AI Project together, each person building one piece of it with AI.
No slides and nothing to install. Everyone opens the same brief on their own laptop and starts, at their own level.
When most of your tests pass, raise a hand. A coworker who is at least as far takes it and sits beside you.
Then you evaluate someone else. Judging another build is half the learning.
Completion isn’t self-declared and there is no trainer to wave you through. You’re done when the person beside you signs.

Pick your seats, invite your coworkers, and we’ll scope the day with you.
One full day, or two half days · 10 to 50 coworkers