I’ve been noticing a subtle but significant shift in PM job descriptions and in my conversations with peers lately. The old debate was, “Do PMs need to code?” The new, more pressing question seems to be, “Do PMs need to understand data science?”
With AI and machine learning moving from a niche feature to a core component of many products, the ground is moving under our feet. Simply being the “voice of the user” is no longer enough when the product experience is driven by non-deterministic models. Understanding the basics of training data, model evaluation, and the user experience of probability is becoming table stakes.
If we can’t grasp the fundamentals of what our data science and ML engineering counterparts are building, how can we effectively challenge assumptions, define success metrics beyond simple accuracy, or communicate the ‘why’ behind a product that sometimes behaves in unpredictable ways? We risk becoming backlog managers, unable to truly shape the product’s intelligence and its interaction with the user. This isn’t about building the models ourselves, but about having the literacy to ask the right questions and guide the product with a deeper understanding.
For PMs working on AI-powered products, what specific data science or ML concepts have you found most crucial to learn, and how has it changed your collaboration with your technical teams?