Product management was built on a simple assumption: We know what users want.
Interview them. Survey them. Watch them use your product. Analyze their behavior. Do all of this and you'll understand their needs. Then build toward that understanding.
It's a sensible framework. It works 70% of the time.
But what happens in that remaining 30%? What happens when users don't know what they want? When they can't articulate it? When what they think they need and what they actually need are completely different?
This is where AI changes everything.
The Blind Spot in Traditional Research
User research is fundamentally constrained. Users can only tell you what they're aware of wanting. They can only report on needs they consciously experience.
But most human needs are unconscious.
You don't know you need better lighting until you experience poor lighting. You don't know you need faster load times until you've waited for slow ones. You don't know you need a feature until someone shows it to you.
Traditional product development asks users: "What do you want?"
Users answer based on their current constraints and awareness. Then we build toward that answer.
But AI can see patterns we can't. It can watch millions of user interactions and identify behaviors that users themselves don't consciously notice. It can reveal latent needs—things users want but have never articulated.
This is the shift: From "What users say they want" to "What users actually need."
The Problem of Infinite Choice
In a world where AI can generate infinite features, test them instantly, and measure their impact, we face a new problem: How do we choose what to build when everything works?
You can generate 100 feature variations in an afternoon. Run them all as experiments. See which one converts best.
But just because something works doesn't mean it's what we should build. Just because users engage with a feature doesn't mean it's meaningful. Just because something is optimized doesn't mean it's right.
This is the paradox of optimization: The more tools we have to predict and measure, the less certain we become about what actually matters.
We can measure engagement. We can't measure meaning.
We can optimize for metrics. We can't optimize for mattering.
Building for What's Possible vs. What's Necessary
There's a distinction that matters here.
What's Possible: All the features AI can help us build. All the optimizations we can run. All the variations we can test. The answer is infinite.
What's Necessary: The features that genuinely improve human life. The optimizations that serve a purpose beyond conversion. The variations that matter because they solve real problems or create real value.
Traditional product management asks: "What's possible that users want?"
The new question is: "What's necessary that users don't even know they need?"
This requires a different kind of thinking. Not analytical thinking. Philosophical thinking.
It requires you to ask:
- What is this product for?
- What is the human need beneath the surface request?
- What will this feature do to the user's life, not just their engagement metrics?
- What am I optimizing for—their happiness or their addiction?
- If I build this, what kind of person am I encouraging them to become?
These aren't data questions. They're wisdom questions.
The Flip from Data to Conviction
As AI becomes better at reading data and predicting behavior, the competitive advantage stops being "I can read the data faster."
It becomes "I can see what the data is missing."
AI will tell you what users want based on their past behavior. But it can't tell you what users need to become better versions of themselves. It can't build products that make people brave, or wise, or more connected to meaning.
Those require conviction.
They require you to build not based on what the data says, but based on what you believe about human potential.
Steve Jobs didn't ask users what they wanted. He imagined what they could have and built toward that vision. The data didn't predict the iPhone. Conviction did.
In a world of perfect data and AI optimization, conviction becomes scarce. And scarce things become valuable.
Building in Genuine Uncertainty
The irony is that despite (or because of) all this AI capability, we're actually building in greater uncertainty than ever before.
When there was less data, you had fewer choices. Fewer variations. Fewer possibilities. The path forward was clearer because the options were fewer.
Now, with infinite options, the path is murkier. You can optimize for anything. You can build anything. You can test anything.
But which direction are you actually walking in?
This is why the best products in the next decade won't be the ones built on the most data or the best AI.
They'll be the ones built with the clearest sense of purpose.
Products built by people who know not just what's possible, but what's necessary.
People who can look at infinite options and say: "This is what matters. This is what I'm building."
Not because the data says so.
But because they believe so.