Why cultivating agency matters more than cultivating skills in the AI era | Max Schoening (Head of Product, Notion)

Max Schoening argues that AI changes the relative importance of skills and agency. Skills can increasingly be summoned through capable models, but agency determines whether someone treats the surrounding world as changeable and acts without waiting for a fixed role or permission. He connects this to the observation that a project's first 10 percent is becoming nearly free, making it easier to build an initial startup version and test an idea rather than merely describe it.

At Notion, the practical route into AI-assisted product work was a small, LLM-friendly playground for chat prototypes rather than an immediate attempt to change the entire production codebase. This reduced fear of the terminal and let designers and product managers experience the material directly. As model capabilities improved, the same people began contributing to production code to a limited degree. Schoening says the point of coding is not necessarily that a designer's code ships, but that working in the material teaches the person what agent loops and AI-mediated interfaces can do.

Schoening distinguishes this approach from indiscriminate feature delivery. He is more interested in designers and PMs understanding the medium than in making them part of a faster delivery mechanism. He also warns that role convergence can remove specialists. His hardware analogy separates rough 3D-printed prototypes from the engineering needed to manufacture a reliable product at the scale of hundreds of millions of people. In software, he sees too much attention on token consumption and feature volume and not enough on reliability, precision, and the work required to make systems operate at scale.

The conversation treats taste as a trainable form of prediction rather than an unexplained gift. Schoening describes it as mentally simulating how a defined in-group will respond to an idea. Repeated attempts, feedback, side projects, exposure to other people's work, and contact with well-made objects all provide the repetitions that improve that simulation. This account also limits the claim that taste is uniquely human: he compares the idea-and-feedback loop to model training.

The early exchange also establishes Schoening's product thesis: strong products usually have a very small core capability that feels like a superpower. Adding features to compensate for a weak core creates a loop that does not resolve the underlying problem. His claim about knowledge work as an existing form of universal basic income is presented as a deliberately provocative observation, not as a detailed economic proposal.