"Every product manager will need to get comfortable with having a partner that's a research scientist. This is not ChatGPT — it's actual PhD researchers building models into your product."
The future has AI baked into every product by default
Personalization, recommendations, automation require ML expertise
PMs must learn to work with research scientists, not fear them
The mindset shift: embrace uncertainty in research timelines
Don't Do AI
The Shiny Object Trap
0
ROI from shiny object trap
Problem-first approach: Is there pain point? Real user need?
No MVP with models: Don't waste data scientists' time training on an unvalidated idea
Use existing data: Leverage adjacent product data or synthetic data first
Define success clearly: What does "solving the problem" actually look like?
Marily's definition
"The generalist PM helps their team build the right product. The AI PM helps their team solve the right problem."
ChatGPT as Your Assistant
AI Tools PMs Use Today
Mission Statements: Feed ChatGPT your draft; it makes them universally understood by execs, juniors, everyone
User Segments: Ask it for personas it'd never occur to you to imagine, complete with motivations & pain points
Feature Ideation: Generates novel AI-enhanced feature ideas for your existing product
Boilerplate Copy: Frees you from tedious writing to focus on strategy
The prompt approach
"Rewrite this mission statement to be understood by anyone, from kids to C-suite." Start with your thinking, let AI polish it.
Not replacement
These tools amplify your judgment. You're still the PM. You direct the work. Tools make you faster and smarter.
Playbook
Getting Started: PM in AI
Learn the fundamentals: take a coding course (Coursera, Stanford intro to AI)
Shadow ML researchers on your team 1 hour/week to understand what they do
Start with data you already have: what existing product data can you leverage?
Think problem-first: "What user pain point could AI solve?" not "How do I use AI?"
Pair with a data scientist to scope a small proof-of-concept, not an MVP
Learn to code
Career Foundry, General Assembly, Coding Dojo offer bootcamp-style coding education. Even a few weeks gives you confidence to talk to ML experts.
Contrarian
AI PM Myths Vs Reality
✗ChatGPT will replace PMsINSTEAD →✓ It frees PMs to do strategic work. Tools amplify, not replace, human judgment.
✗AI is only for big techINSTEAD →✓ Any PM with user data can identify an AI opportunity: fraud, personalization, automation, recommendations.
✗You need massive datasets to startINSTEAD →✓ Simple models work with hundreds of examples. Start small, prove the concept, scale the data.
✗You need to code to be an AI PMINSTEAD →✓ Learning fundamentals builds confidence. You'll never train models, but you'll speak the language.