Why Product Managers Who Cannot Read Data Will Struggle to Lead in the Age of AI – Chinwendu Nwaozuzu

At an informal product meetup last year, I found myself in a conversation that I have been turning over ever since. The discussion centred on how reliant product managers can be on the data team. The prevailing view was that this reflects a reasonable division of labour. I understand why. Product management is already an impossibly broad role, and there is logic in working with specialists.

But I kept thinking about it—because the world that logic was designed for is changing faster than most organisations have noticed.

Cassie Kozyrkov, who built and led Google’s decision intelligence function, has made a version of this argument with more rigour than I can claim: the most consequential decisions in any data-rich environment are not made by the people who run the analyses; they are made by the people who frame the questions. Increasingly, in a world where AI handles more of the analytical layer, the ability to frame the right question—to know what to ask of your data and why—is becoming the defining skill in technology leadership.

Product managers who treat data as a support function they receive, rather than a language they speak, are, I would argue, quietly ceding their most important capability.

I did not arrive at this view through theory. I arrived at it through reflection on a GCP data migration I led as interim project manager. A critical pipeline failure emerged mid-project. The engineering team had identified a fix, but it carried a risk of temporary data inconsistency. The alternative was to pause and absorb a delay that would have consequences far beyond the project itself.

There was no time for a considered analytical process—barely time for a conversation. I had to assess the technical risk, weigh it against the delivery risk, make a decision, and communicate the reasoning to stakeholders clearly enough to maintain their confidence, all within a window measured in hours.

I chose to proceed. The migration was delivered on time. But what I remember most clearly is not the outcome; it is the feeling of understanding the data well enough to know what I did not know—and being confident enough in that distinction to act.

That is what data fluency actually means in practice. It is not the ability to write SQL or build a regression model. It is the ability to look at a dataset and ask the right questions. To understand what a metric is measuring—and what it is not. To recognise the difference between a correlation that matters and one that simply flatters your assumptions.

Over the course of my career, the question I have most often heard from product managers at various stages is some version of: Do I need to be technical to be a great PM? The framing of that question is revealing. It treats technical knowledge as a threshold to either clear or not, rather than as a spectrum of capability to develop continuously.

My answer has been consistent: you do not need to be an engineer, but you do need to be able to have an honest conversation with one. You need to understand what your data is telling you well enough to push back when interpretations do not align with user realities. And as AI tools become more deeply embedded in product workflows—generating insights, suggesting prioritisation, synthesising user feedback at scale—the PM who cannot interrogate those outputs is not being helped by AI; they are being led by it.

This matters more than it might appear. AI systems are extraordinarily good at finding patterns. They are also, as anyone who works with them closely knows, remarkably good at finding plausible-sounding patterns that do not hold up. The capacity to distinguish between a genuine signal and an artefact of the model is not something AI can reliably do for itself. That judgement either sits with the human in the loop—or it sits nowhere.

There is a broader argument here about the kind of leadership technology organisations actually need. DJ Patil, the former Chief Data Scientist of the United States, has described the ideal data-driven organisation as one where data literacy is not confined to a specialist function but distributed across every layer of decision-making. Insight should sit as close as possible to the person making the call. The further apart those two things are, the more signal gets lost in translation.

Being shortlisted for the TechWomen100 Awards brought me into contact with an extraordinary range of women doing transformative work across the technology sector. What stood out, across almost every conversation, was that the leaders moving fastest were not those with the deepest single specialism. They were the ones who could move fluently between layers—from data to user insight to strategic decision-making—without losing fidelity at any point.

That fluency is not an innate talent. It is cultivated deliberately, over time.

The product managers who will lead in the next decade are not necessarily those who can code or build models. They are the ones who have taken the time to understand what their data is doing—and who have the confidence to question it when something does not feel right.

That instinct—the willingness to interrogate outputs rather than simply consume them—is, I would argue, the most important leadership quality the age of AI demands. And it is one that no AI can develop for you.