"AI product sense is learned through a weekly practice of provoking failure modes, predefining quality and cost thresholds, and designing explicit guardrails for uncertainty."
Evidence from the Archive
Meta / Google
Deciding whether to launch a model at 70% or 80% accuracy -- a product decision that requires technical understanding
The advice to 'fake the AI with a Figma prototype' for MVPs instead of investing in model training prematurely
Marily Nika has led AI product management at both Google (AR/VR team) and Meta, making her one of the most experienced practitioners of AI PM work at scale -- her advice comes from years of navigating the specific challenges of shipping ML-powered products with research scientists. Their core argument: PMs should absolutely learn to code -- it is the bridge to working effectively with researchers and making sound AI product decisions.
The evidence is specific: Deciding whether to launch a model at 70% or 80% accuracy -- a product decision that requires technical understanding. Furthermore, the advice to 'fake the AI with a Figma prototype' for MVPs instead of investing in model training prematurely. Working with Google's AR/VR team where research scientists needed a PM who could engage with model uncertainty.
In Marily Nika's own words: "I believe that ballpark managers will be AI product managers in the future. And this is because we see all products needing to have a personalized experience, a recommender system that is actually good." (On why every PM will eventually be an AI PM.)