How Notion uses AI to increase product manager autonomy (and you can do the same)

The pressure to adopt AI in product teams is everywhere. But the conversation usually starts in the wrong place: which tool to use, how to automate tasks, where to deploy a copilot.

Max Schoening, Head of Product at Notion, proposes flipping the logic. In a recent interview on Lenny’s Podcast, he argues that what sets a PM apart in the AI era isn’t mastering the tool — it’s having agency. The ability to identify what needs to be done and act to solve it, even without all the technical skills required.

This changes how you structure teams, how you hire, how you develop people. And you don’t need to work at big tech to apply it.

What is agency (and why it matters more than skills)

Agency, in the context of product, is the willingness to solve a problem without waiting for permission, a process, or the perfect tool. It’s the opposite of passivity.

Schoening makes a useful distinction: skills are what you know how to do. Agency is what you do with what you know, or what you figure out how to do when you don’t know.

Generative AI tools let a PM write SQL, generate prototypes, analyze data, create documentation. But none of that has value if the person doesn’t know which problem to attack, how to prioritize, or when to stop exploring and decide.

It’s common in the market to see teams investing in technical training for PMs (SQL, Python, machine learning basics) while ignoring the development of autonomy and judgment. The result is professionals who can operate tools but wait for someone to tell them what to do with them.

How Notion structures this in practice

According to Schoening, Notion’s model for product teams is built on a few principles:

Context over control. Leadership spends more time sharing strategic context than defining tasks. The expectation is that the PM, with enough context, knows what to do. If they don’t, the problem is context, not capacity.

Internal tools that amplify, not replace. Notion builds internal AI tools that help PMs run analyses faster, but the judgment about what to analyze and what to conclude stays human.

Hiring for agency, not tool resume. In recruiting, the question isn’t “do you know how to use X”. It’s “tell me about a time you solved a problem without having ideal conditions.”

Common approach

• Train PMs on AI tools
• Define tasks and measure execution
• Hire for technical skills
• AI as a productivity solution

Notion approach

• Give PMs context to decide
• Share strategy and expect initiative
• Hire for demonstrated agency
• AI as a judgment amplifier

Why this matters for smaller teams

You might think: “this works at Notion because they have resources, time, and room to fail.”

Schoening’s argument goes the opposite direction. The smaller the team, the more agency matters. In a lean structure, there’s no space to wait for approval from three levels up or a perfect process. Whoever acts, moves forward. Whoever waits, gets stuck.

AI only amplifies this. With generative tools, a PM with agency can do in an afternoon what once took days of alignment and support from other departments. But that only happens if the person is clear on what needs to be done and willing to act.

The risk, of course, is acting without direction. That’s why the counterpart is investing in context. The more autonomy you give, the more strategic context you need to share. It’s not cutting people loose without support. It’s equipping them to decide well.

How to develop agency on your team

If you lead product or work on a team that wants to adopt this logic, a few questions help diagnose where the bottlenecks are:

• Do people on the team know what problem we’re trying to solve this quarter?
• Do they have access to the data they need to make decisions?
• When someone proposes a solution, is the default to approve or to push back?
• Is there room to fail on small things without escalating to leadership?
• Was the last hire evaluated on initiative or only on prior experience?

If most answers point to centralized control, the problem isn’t lack of technical skill. It’s lack of structure for agency.

The path isn’t “give total freedom all at once.” It’s expanding the surface area of decision-making gradually, observing how people respond, and adjusting the level of shared context as needed.

What this changes about AI adoption

AI adoption in product teams usually starts with the tool: “let’s use Copilot,” “let’s test Claude for documentation,” “let’s integrate AI into the backlog.”

The model Schoening describes suggests starting from the other side: before you pick the tool, ask if the team has enough clarity to know what to do with it.

Generative AI is a multiplier. If the team already has agency, the tool accelerates. If the team operates in “wait for instruction” mode, the tool becomes another underutilized thing sitting in your stack.

This doesn’t mean ignoring technical training. It means flipping the order: first, develop the ability to act with autonomy. Then, give the tools that amplify that ability.

The risk of doing nothing

The product market is dividing into two groups: people using AI to do more with less, and people still operating like nothing changed.

The difference between the two isn’t tool access. It’s posture. The first group has agency: they see AI as an amplifier and act to capture that value. The second group waits for someone to define how to use it.

Schoening’s lesson, at its core, is less about Notion and more about a structural shift in how product work happens. The technical barrier is falling. What’s left as the differentiator is the ability to decide what to do and the willingness to act.

Reference: Max Schoening’s interview on Lenny’s Podcast, episode “Why cultivating agency matters more than cultivating skills in the AI era”.