Why cultivating agency matters more than cultivating skills in the AI era | Max Schoening (Notion)
Max Schoening, Head of Product at Notion, discusses how AI is reshaping software development. He argues that agency—the drive to effect change—is now more critical than specific skills, as AI makes the first 10% of any project 'free'. Schoening introduces his 'tiny core' theory for great products, debates the 'SaaSpocalypse', and emphasizes the growing importance of software quality and taste in an era of exploding software quantity.
The End of Silos: Agency Over Skills
In the age of AI, the traditional lines between product, design, and engineering are blurring. The most crucial attribute for success is no longer a specific skill set but agency—the deep-seated belief that the world around you is malleable and can be changed. Previously, a “skill issue” could be a valid reason for not being able to build something. With AI tools, those barriers are dissolving, revealing that the primary differentiator is the will and drive to act and create.
At Notion, this has manifested in a culture where designers and PMs are encouraged to prototype directly in code. The goal isn’t necessarily for them to ship production code, but for them to understand the medium they are designing for.
“The reason I like thinking in code is because it forces you to consider the medium… It’s really important not to forget that the reason why is to to become a master of the material, not a sort of cog in the delivery mechanism for the idea.”
This hands-on approach allows them to grasp the nuances of modern AI systems, such as agentic loops, which is far more valuable than simply tweaking UI styles. High-agency individuals don’t confine themselves to a job description; they identify what the organization needs and take the initiative to learn the skills or find the people to make it happen.
The New Product Lifecycle: The First 10% is Free
AI has fundamentally altered the economics of starting a project. The initial 10% of any endeavor—the brainstorming, the initial prototype, the first version of a startup—is now effectively free. It takes minimal effort to get a functional demo up and running, which dramatically accelerates the exploration phase.
This shifts the product development process towards a “demos, not memos” culture. Instead of writing lengthy PRDs, teams can build and present multiple tangible versions of a product, allowing for much richer, more immediate feedback. However, this doesn’t eliminate the hard work; the last 10% of a project (refinement, quality, scale) still represents 90% of the effort.
The “Tiny Core” Theory of Great Products
The most successful and enduring products are not the ones with the most features. Instead, they are built around a tiny core—a single, exceptionally well-executed superpower that is almost magical. Adding more features in the hope of achieving greatness is a common pitfall that rarely works.
Examples of a “tiny core”:
- iPhone: Multitouch interface.
- GitHub: The pull request.
- Notion: The block and slash commands.
- Heroku: The
git push heroku mastercommand. - Dropbox: The simple, reliable menu bar icon for syncing.
The focus should be on identifying and perfecting this core, as everything else flows from its strength and simplicity.
Malleable Software and the “SaaSpocalypse”
The concept of malleable software is the idea that software should serve the interests of the people who use it, not just the corporations that create it. Users should have ownership over their computing environment, with the ability to change and adapt tools to their specific needs, much like rearranging a living room.
While there’s talk of a “SaaSpocalypse” where everyone builds their own tools, the “as-a-service” model is unlikely to disappear. Most people don’t want the burden of maintaining the full stack of their software. The value of SaaS is in the maintenance, security, and specialized expertise provided. The future likely involves more general, malleable platforms (like Notion) that act as operating systems for users to build upon, rather than a complete replacement of the service model.
Quality and Taste in an Age of AI-Generated Code
While the quantity of software has exploded in the last year, the quality has not seen a corresponding increase. There is a significant opportunity for products that are reliable, well-crafted, and demonstrate a high level of taste.
- Shipping and Quality: The key is to balance “shots on goal” with a commitment to “incremental correctness.” It’s good to explore many ideas and ship frequently, but it’s crucial to then do the hard work of consolidating those ideas into a cohesive, high-quality, “obviously good” core.
- Developing Taste: Taste isn’t mystical; it’s the ability to run a “virtual machine in your head” to predict how a specific audience will react to an idea. It’s a model trained through countless reps: building, shipping, getting feedback, and iterating. Exposure to other tasteful products is also critical for setting a high bar for your own work.
The Business and Culture of AI
- ROI on Tokens: Currently, many companies are in an exploratory phase, allowing for unlimited token spend to foster innovation. However, within the next 6-12 months, uncomfortable conversations about the ROI of this spend will become common.
- AI Adoption: Getting teams to change their workflows requires significant effort. For roles further from engineering, the leap in capability is so intoxicating that adoption is easy. For engineers, a good litmus test is to treat any manual human intervention in the coding process (aside from code review) as a bug in the “software factory.”
- Future of Roles: AI’s primary impact will continue to be in accelerating software development. Rather than AI transforming other domains directly, the cost of software will approach zero, leading to software engineering principles “eating” all other domains as they become automated and encoded.