Marty Cagan and Marily Nika outline what makes AI product management distinct from conventional product management, framing it through the lens of four critical product risks: feasibility, usability, value, and viability. They argue that just as "Mobile PM" was once a specialized skill that eventually became table stakes, AI product management will soon be expected of all product managers. The article focuses specifically on AI-powered applications rather than underlying infrastructure, drawing a parallel to the platform-versus-experience product distinction.

The authors detail how each risk category intensifies with AI. Feasibility risk is heightened because generative AI is probabilistic, not deterministic — making training data quality, error rates, and appropriate use-case matching paramount. Usability risk demands new approaches to transparency, explainability, and trust-building in the user experience. Value risk requires product managers to ensure AI features deliver genuine incremental benefit rather than serving as marketing checkboxes. Viability risk encompasses unit economics (still uncertain for many AI products), data provenance and copyright concerns, legal liability for probabilistic outputs, and ethical considerations around bias, hallucinations, and misuse by bad actors. The piece underscores that these viability questions fall squarely on the AI product manager's shoulders, making deep knowledge of users, data, business context, and technology more critical than ever.

Marty Cagan of Silicon Valley Product Group reflects on two years of generative AI's impact on product management, distinguishing between AI as a feature in products versus AI as a tool that changes how products are built. He acknowledges that despite a flood of announcements and bold predictions, the actual transformation has been slower than he hoped — his end-of-2023 predictions proved too optimistic about the pace of change.

Cagan highlights that most industry discussion focuses on delivery team product owners or feature team product managers, rather than empowered team product management, which is his primary area of interest. He notes that while AI tools clearly accelerate code creation and design work, faster production doesn't necessarily translate into better outcomes. The article poses several unresolved questions for product model companies, including how generative AI changes product discovery, which PM skills become more or less important, and whether AI will drive genuine innovation or merely incremental optimizations.

A key tension Cagan identifies is the pairing of strong product judgment with AI tools — he sees enormous potential when someone with deep product sense uses generative AI, but worries about the risks when people lacking foundational product skills rely on the same tools. Two years in, he concludes that while real progress has been made, the industry is still at the beginning of understanding AI's role in product management.