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

Max Schoening, Head of Product at Notion, joins Lenny's podcast to discuss the profound shifts happening in product building as AI becomes more powerful. Max has an unusually multidisciplinary background spanning engineering, design, and product management at companies like GitHub, Heroku, and Google, which gives him a unique perspective on where things are heading.

On the topic of role convergence, Max explains how Notion designers and PMs began coding not to ship production code, but to understand the material they design with. He created a simplified 'playground' codebase optimized for LLM-assisted prototyping, especially for chat interfaces, citing Brett Victor's 'Stop Drawing Dead Fish' talk as inspiration. His core argument is that coding forces designers and PMs to deeply understand the medium — particularly agent loops — rather than designing static mockups divorced from reality.

Max argues that the most important differentiator in the AI era is agency, not technical skill. He defines agency as the understanding that the world is malleable and that one can change things — a trait he sees as unevenly distributed. He distinguishes high-agency employees (like Brian Levin and Eric Liu at Notion) as those who redefine their roles based on what needs to happen rather than clinging to job descriptions. He connects this to the Steve Jobs quote about realizing the world is made by people no smarter than you.

On malleable software, Max argues that most software today serves the corporation that makes it rather than the people who use it. He envisions software that adapts to individual users the way a well-lived-in home adapts to its inhabitants over time. He connects Notion's success with AI agents to this philosophy — agents need open, connected environments to operate effectively, and Notion's structure provides that.

Max is skeptical of the 'SaaS apocalypse' narrative, arguing that people don't actually want to maintain their own software stacks. The 'as a service' model persists because it handles maintenance and specialized thinking, which users don't want to do themselves. He predicts tools will become more general (like 90s-era word processors and spreadsheets) but remain as-a-service.

On the future of product building, Max says the 'first 10% of every project is now free' — it's trivially easy to build a v0 prototype — but the last 10% remains 90% of the work. He sees AI expanding the ability to explore multiple product directions simultaneously rather than fundamentally changing the nature of product quality work. He's concerned that vibe coding has increased the volume of software without increasing its quality.

On taste, Max defines it as the ability to run a 'virtual machine in your head' — to predict how a specific in-group will react to an idea. He argues taste is built through repeated iterations with feedback, analogous to how models are trained. He notes this makes him somewhat skeptical that taste is uniquely human and uncopyable by AI.

Max shares his 'hot take' that knowledge work is already a form of universal basic income — that comfortable, air-conditioned cognitive labor is far more than people need to live, and that humans will always find ways to insert themselves into agentic loops regardless. He closes with advice for young people in Silicon Valley to reduce the frenetic anxiety about missing the 'last train' and instead focus on genuine curiosity and tinkering.