Free Lesson
No Vibes, Just Evals: Proven Frameworks for AI-Native PMs
Part of The AI-Native Product Manager
30 min
Mar 20, 2026 4:30 PM
Virtual (Zoom)
In this video
What you'll learn
Define "good" before you write code
Evals are replacing prds. Learn the framework that tests judgement before launch.
Write behavior specs, not feature specs
AI doesn't execute instructions - it interprets them. Learn to encode judgment so your model does what you actually mean
Set your autonomy dial (live demo of real product)
Make decisions to govern judgement - what AI controls and what humans control, using an autonomy map.
Why this topic matters
Al features can look "done" and still fail users in production. AI-Native PMs fix this. They define success criteria, encode judgment into behavior specs, and decide autonomy levels for AI - before incidents happen in production.
This talk gives PMs 3 battle-tested frameworks, the same principles used in industrial eval infrastructures that serve 100M+ users. (includes real product live demo)
You'll learn from

Jason P. Yoong
Co-Founder & COO at Level Up (Prev: Amazon, Dentsu, IPG, Startup - $8M seed)

Shankha Dey
Sr Director PM @ Salesforce | ex-Microsoft Bing & Copilot | Columbia CS
Senior leaders and advisors at
