Build the Harness: A Hands-On Challenge in AI Agent Reliability

A prompt and a model alone don’t make an agent you can rely on in a company. What makes it accurate, secure and trustworthy is the harness around it, and getting that right takes more decisions than most teams expect.

In this hands-on challenge, you compete solo or as a team. You get a set of example tasks from a corporate setting and build an agent to solve them, running on real AI models with a token budget. Every attempt gets scored, and the highest score wins.

Then we look behind the scoreboard: how the agents performed on inputs they had never seen, which approaches held up, which ones broke, and what a genuinely strong solution looks like.

You leave with a practical understanding of what it takes to build AI agents that work reliably in a real company.

No coding and no preparation needed, AI tokens are provided. Bring a laptop, at least one per team.