What Leading AI Products Taught Me About Product Strategy
What Leading AI Products Taught Me About Product Strategy
One of the biggest surprises of working on AI products is that many of the product management principles I learned over the last decade still apply.
But one of them changes completely: Predictability.
In traditional software products, there is usually a relatively clear relationship between effort and outcome.
You define requirements.
Engineers build the solution.
Users interact with it.
You measure the impact.
The path is rarely linear, but it is generally predictable.
AI products are different.
Over the last few years, while leading AI initiatives, I've learned that product strategy becomes much more about managing uncertainty than defining certainty.
The reason is simple: AI operates in a non-deterministic environment.
You can have a promising use case, a strong business problem, quality data, and talented engineers, and still discover that the model doesn't produce results good enough for real-world adoption.
Unlike traditional software, where the question is often "Can we build it?", AI introduces a different question:
"Will it work well enough?" And the honest answer is often:
"We don't know yet."
This uncertainty fundamentally changes how product strategy should be approached.
Instead of committing to large roadmaps upfront, I've found it more effective to think in terms of learning milestones rather than delivery milestones.
The objective is not to build a feature. The objective is to reduce uncertainty.
Can the model achieve the required accuracy?
Can users trust the outputs?
Does it create measurable business value?
Will people actually change their behavior because of it?
Each experiment is an investment in knowledge. Sometimes the result is a successful product. Sometimes the result is discovering that the problem isn't worth solving with AI at all. And that's valuable too.
The biggest lesson AI product management has taught me is that strategy is not about pretending we know the answers. It's about creating a systematic way to discover them.
In a world where outcomes are uncertain, learning becomes the most important metric.