The Punk PM #41

Hey there, punk!

It's half-term this week, so I've been juggling work and childcare with my usual grace and composure. Translation: I haven't had much time to think.

But something's been bubbling away in the background: how much AI is disrupting the ways you "win" in tech, and how stubbornly we're all clinging to the old patterns.

Let's get into it.

Quote of the Week 🙊

Insight 🦉

There's a pattern that's defined some of the biggest successes in the last twenty years of tech. Find a layer that others have to build on top of and own it.

Stripe owned payments. Salesforce owned the customer relationship. Google owned search. Get embedded deep enough in the stack, become infrastructure, and collect rent.

It worked because distribution was hard. Building the pipes was the moat.

LLMs are breaking that in two directions at once. From below, agents can now operate across tools on behalf of a user which makes the UI layer and the distribution layer both matter less. The pipes still exist, but the AI sits above them, abstracting them away. From above, something like ChatGPT can capture user intent before they ever reach your marketplace or your SaaS product. Lose the starting point of the customer journey and you lose the relationship, the data, and eventually the pricing power.

But here's the interesting thing. The companies building AI are largely recreating the exact patterns they're disrupting. Building horizontal platforms. Racing to become infrastructure. Converging on the chat box the same way every SaaS company converged on the dashboard.

That's not stupidity, it's structural. The incentive systems push you toward the incumbent playbook even as you're theoretically tearing it down. You end up optimising for how value was captured in the old world, not the new one.

So what does a genuine new moat look like? The most likely is stuff that's hard for AI to aggregate. Depth and exclusivity over breadth. If an AI can commoditise anything that can be described and retrieved, the defensible thing is what can't be. The leverage is no longer in building infrastructure. It's in understanding a workflow deeply enough to automate it in a way that's specific, not generic.

The question isn't whether the game has changed. It's whether you can see it clearly enough to do anything about it.

Action 🚀

Next week, pick one product you use every day and write one sentence: what workflow does it own, and why would that be hard for an AI agent to replace?

Then ask yourself if you could answer the same question for your own product.

Inspiration 💡

LLMs vs Marketplaces – Dan Hockenmaier makes a sharp case for which businesses are most vulnerable to AI disruption, and which ones aren't. The signal: if your supply is easy to aggregate, you're exposed. If you're deeply embedded in the workflow, you might be fine. For now. Read more

Everyone in AI is Building the Wrong Thing for the Same Reason – Joan Westenberg on why AI companies keep copying each other into irrelevance. The incentive systems are the problem, and escaping them requires staying small, staying focused, and resisting the pull of the obvious. Read more

How Generative and Agentic AI Shift Concern from Technical Debt to Cognitive Debt – Margaret-Anne Storey introduces a concept worth sitting with. AI can generate code faster than teams can understand it, and that gap between shipping and knowing is where the real risk lives. Read more

Signing Off ✍️

If this resonates with you, hit reply and let me know. And if you think a friend or colleague would enjoy The Punk PM, feel free to share it with them!

Play it your way,

Toby