When Shipping Gets Cheap, Focus Gets Expensive
The most dangerous thing AI did to product management wasn’t replace jobs. It removed the one thing that used to force good decisions: the cost of saying yes.
The Old World Had a Hidden Feature
Building software used to be slow. Not just slow, but also expensive in the specific way that made everyone in the room think twice before committing.
You needed engineering time. That meant trade-offs. You needed design cycles. That meant priorities. You needed stakeholder sign-off. That meant explaining, defending, refining. And in large organizations, you needed heavy governance: the frustrating, sometimes paralyzing layer of accountability that made everyone ask the same question before anything got built. Is this actually worth it?
Those systems were maddening. They were also clarifying.
Call it the scarcity dividend.
When saying yes is genuinely costly, you get better at saying no. You learn to know where you’re going before you start. That discipline, forced on you by scarcity, produced depth. It produced things that lasted.
What Actually Changed
The conversation about AI in product tends to focus on speed. You can ship in days what used to take months. That’s true, and it matters.
But speed isn’t the thing that changed. Consequence is.
When yes is nearly free, the hardest product skill, choosing, quietly disappears. You’re not building faster. You’re avoiding the question faster. The question being: what are we actually trying to be?
The Failure Mode
You’ve watched it happen. Maybe you’ve lived it.
Someone has a customer conversation. A founder, a sales rep, customer success, it doesn’t matter who. A new feature comes up. It sounds reasonable. There’s no sprint planning meeting to negotiate, no engineering backlog to fight through. Anywhere from hours to a few days later, it’s in the product.
Then another conversation. Then another. The surface area expands. Each addition makes sense in isolation. The aggregate is a product that does many things adequately and nothing the market can’t live without.
No moat. No clear identity. No specific user who picks it up and thinks: this was built for me. Just a sales deck that takes twelve slides to explain what the product actually does.
The product doesn’t fail because the features are bad. It fails because there are too many of them, and none of them are deep enough to matter.
This Isn’t a Founder Problem
An IC PM at a 500-person company can spread a product just as thin as a solo founder with unlimited AI credits. The disease is the same. And the moment you’re most excited about a new idea is the moment your judgment is least reliable.
Excitement is a signal.
It just doesn’t always point where you think it does.
Restraint Is a Product Decision, Not a Resource Decision
There’s a version of this story that ends with: slow down, be thoughtful, resist the urge. That’s not what I’m saying.
The answer isn’t to artificially reintroduce friction. It’s to understand what friction was actually doing for you, and replace it with something intentional.
The goal is still what it always was: world-class at one thing, for a specific user, in a specific context. That’s still the only thing that builds a moat. Real feedback from that specific user is still the only signal worth optimizing for.
AI gets you there faster. But only if you know where “there” is before you start.
Restraint Needs a System
Knowing this isn’t enough. I’ve watched people who understood it completely still drift, because the pull of a new idea is strong and the cost of chasing it feels low.
Restraint needs a system. Not a gate. Not a checklist. A thought partner that holds your true north when your judgment is at its worst: when you’re excited, when a customer conversation just ended, when the idea feels obvious and the build feels easy.
I built something for this. It’s called Anchor.
Before anything gets built, it establishes three things: who the target user actually is, what winning looks like for them in observable terms, and the one thing the product has to do better than anyone else. That becomes the constitution. Every new feature gets interrogated against it. Not to block it. To surface the tension before you commit.
You leave the conversation more convicted or more relieved. Either outcome is a win.
The rarest skill in the AI-native product era isn’t prompting or shipping speed. It’s being able to finish this sentence: we build this, for these people, until it does this one thing better than anyone else.
Anthony Russo writes about navigating the gap between where markets are and where they need to go. Building Through Friction publishes when there’s something worth saying.