Why cultivating agency matters more than cultivating skills in the AI era | Max

Max Schoening's observation hits different when you've watched talented PMs get outmaneuvered by mediocre ones with better instincts. The Head of Product at Notion isn't wrong — the game has changed. While most product managers frantically upskill on AI frameworks and prompt engineering, the ones actually winning are developing something harder to replicate: agency.

I've seen this play out across three different companies over the past eighteen months. The PMs who thrived weren't necessarily the ones who could write the best ChatGPT prompts or build the slickest automation. They were the ones who saw AI as leverage for decisions they were already positioned to make well.

What Agency Means for Product Managers in the AI Era

Agency isn't just decision-making authority — though that's part of it. It's the capacity to identify what needs to happen and the conviction to drive it forward, even when the path isn't obvious. In an AI-accelerated world, this matters exponentially more than technical proficiency.

Think about what's actually happening when a PM "uses AI effectively." They're not just prompting a language model. They're identifying which problems are worth solving, framing questions that generate useful answers, and synthesizing outputs into decisions that move the business forward. The AI amplifies their judgment — it doesn't replace it.

I watched one PM at a fintech startup use Claude to analyze six months of customer support tickets in an afternoon. Impressive, right? Except the analysis sat in Slack for two weeks because she couldn't connect the insights to any decision the team was actually positioned to make. Meanwhile, her peer took the same output and immediately restructured the onboarding flow because he understood which parts of the user journey the team could actually influence.

The difference wasn't AI skills. It was agency — understanding the business context well enough to turn information into action.

Why Traditional PM Skills Are Becoming Commoditized

The uncomfortable truth is that AI is commoditizing much of what we traditionally considered PM craft. Market research, competitive analysis, user story writing, roadmap formatting — these used to be differentiating skills. Now they're table stakes that anyone can execute with the right prompt.

But this isn't necessarily bad news. It's forcing us to confront what product management actually is at its core: making good decisions with incomplete information in service of business outcomes.

The PMs who built their careers on being the person who could synthesize complex data or write crisp requirements are feeling the pressure. AI can synthesize faster and write more consistently. But AI can't decide which customer segment to prioritize when revenue and strategic positioning pull in different directions. It can't read between the lines when engineering says "two weeks" but their body language says "this is harder than we thought."

Consider the explosion of AI-generated PRDs and strategy docs flooding Slack channels everywhere. Most are well-formatted and comprehensive. Few drive meaningful decisions. The skill of document creation has been democratized, but the judgment of what belongs in the document — and more importantly, what gets acted on afterward — remains distinctly human.

How to Build Product Management Agency Over Pure Skill Acquisition

Building agency requires a different approach than building skills. Skills are about capability — agency is about positioning and judgment.

Develop business literacy beyond your product area. The PMs with the most agency understand how their product decisions connect to business outcomes three levels up. They can translate feature conversations into revenue conversations and strategic conversations into resource allocation decisions. When AI surfaces insights about user behavior, they immediately know whether those insights matter to the CFO, the head of sales, or the engineering leadership team.

Cultivate decision-making reps, not just analytical reps. Most PMs practice analysis — gathering data, creating frameworks, building consensus. Fewer practice actual decision-making. Start small: pick one decision per week that you could analyze indefinitely and just make it. See what happens. Build comfort with being wrong and adjusting quickly rather than being comprehensive and slow.

Build context that AI can't replicate. Attend the sales calls. Sit in on customer support escalations. Understand which engineering initiatives your CTO actually cares about versus which ones they're politically obligated to support. This context becomes the filter that makes AI output useful instead of overwhelming.

Focus on the meta-question. When everyone else is asking "How do we build this feature?", agency-driven PMs ask "Should we build this feature?" When teams debate implementation approaches, they're questioning whether the problem is worth solving at all. AI can answer the "how" faster than ever — but it still needs humans to frame the "whether" and "why."

Real-World Examples: Agency-Driven PMs Who Thrived with AI

The best case study I've seen comes from a PM at a B2B SaaS company who used GPT-4 to automate competitive intelligence gathering. Instead of spending hours manually tracking competitor feature releases, she had AI monitor and summarize changes across fifteen competitor products weekly.

But here's what made it powerful: she used that time savings to build relationships with enterprise sales reps and customer success managers. When the AI surfaced that three competitors had launched similar workflow automation features in the same month, she already knew from her sales conversations that this was happening because a major industry conference had highlighted this specific pain point.

She didn't just have the data — she had the context to act on it. Her team shipped a competitive response in six weeks instead of six months because she'd already done the relationship work to understand what customers actually needed versus what competitors were building.

Another example: a growth PM who automated most of his experiment analysis using Claude. Rather than spending Tuesday afternoons in spreadsheets, he used that time to shadow customer onboarding calls. When AI analysis showed that users who completed a specific action in their first session had 40% better retention, he immediately knew which onboarding steps were actually confusing users versus which ones just looked confusing in the data.

The pattern is consistent: AI handles the analytical heavy lifting, agency-driven PMs invest the time savings in building context and relationships that make those analyses actionable.

Sometimes this means better communication too. When your team has clear alignment on strategy and priorities, AI-generated insights can be immediately evaluated against shared context rather than requiring another round of stakeholder education. The narrative clarity becomes the foundation that makes AI assistance genuinely useful rather than just impressively formatted noise.

The product managers who win in the AI era won't be the ones who can prompt engineer their way to better requirements. They'll be the ones who know which requirements matter and have positioned themselves to act on what they learn.