Unlike traditional deterministic products, AI models are probabilistic; the same query can yield different results. This uncertainty requires designers, PMs, and engineers to align on flexible expectations rather than fixed workflows, fundamentally changing the nature of collaboration.

As AI tools automate coding and prototyping, the product manager's core function is no longer detailed specification writing. Instead, their value multiplies in judging, facilitating, and making the right strategic decisions quickly. The emphasis moves from the 'how' of building to the 'what' and 'why,' making decision-making the critical skill.

Building non-deterministic AI products fundamentally changes the PM role. Instead of creating detailed, rigid specifications, the PM's primary task becomes defining and codifying "what good looks like." This is done by repeatedly grading AI outputs to train evaluation systems and guide the model's behavior.

In the fast-moving AI space, long-term roadmaps are obsolete. Anthropic uses lightweight monthly planning for execution and creates 3-6 month vision prototypes—not static decks—to provide directional alignment without creating a rigid plan that will quickly become outdated.

Anthropic's product managers on the research team spec out requirements for each new AI model, defining what it should be good at (e.g., coding, knowledge work). This product development discipline is applied to the inherently unpredictable process of "growing" a model, bridging the gap between research and user needs.

Unlike traditional software, AI products are evolving systems. The role of an AI PM shifts from defining fixed specifications to managing uncertainty, bias, and trust. The focus is on creating feedback loops for continuous improvement and establishing guardrails for model behavior post-launch.

The traditional PM function, which builds sequential, multi-month roadmaps based on customer feedback, is ill-suited for AI. With core capabilities evolving weekly, AI companies must embed research teams directly with customer-facing teams to stay agile, rendering the classic PM role ineffective.

OpenAI operates with a "truly bottoms-up" structure because it's impossible to create rigid long-term plans when model capabilities are advancing unpredictably. They aim fuzzily at a 1-year+ horizon but rely on empirical, rapid experimentation for short-term product development, embracing the uncertainty.

Contrary to assumption, the design process at OpenAI isn't about planning for a distant future. It's a fast-paced environment where designers work in close concert with the latest research advancements, adapting to new capabilities as they emerge.