Designing AI Agent Behavior — A Workshop for Product Teams
AI agents are showing up in products everywhere, but the part nobody talks about enough is this:
Product teams are the ones shaping how these agents actually behave.
Not engineering.
Not data science.
Product.
PMs are now defining tone, boundaries, escalation rules, “I don’t know” states, decision patterns, and what “good behavior” even means. It’s a new responsibility without much guidance — and most teams are quietly figuring it out on the fly.
This workshop is a chance to dig into that work in a practical, approachable way.
We’ll explore:
• How product teams define agent behavior
Tone, constraints, personality, safety, failure states
• How to test and evaluate an agent before launch
What to look for, how to run quick evaluations, how to spot subtle issues early
• How to iterate and version behavior safely
Lightweight workflows PMs can use without diving into ML
• What “agent UX” looks like in real products
Where users get confused, what earns trust, what breaks trust
This is not a technical session. No code, no ML theory.
Just clear mental models, real examples, and space to talk through the challenges PMs are facing as agents become real product surfaces.
Whether you’re already building with agents or just exploring how they fit into your roadmap, you’re very welcome here.