Shubha Govil is Chief Product Officer at Sauce Labs, where she leads product strategy for software testing infrastructure.
More than a decade ago, I was working as a product manager to solve the problem of bringing remote teams together to collaborate in a fully immersive video conferencing environment.
During this project, I camped out in the customer’s office for hours. I spent the better part of the day in a hallway, watching their leaders take video meetings through a glass door.
The feedback we were getting was that the quality of conferencing was low, but the real customer issue I found was about the quality of the screen share, as the drawings presented on the call were in CAD software. What they needed was a content stream sharp enough to carry a drawing across a room without losing the precision of a single line.
Nobody had asked for it that way, so I could only learn that from watching users in their real environment. To solve this problem, we added the high-resolution content share channel. The leader I had been watching told me, years later, that the change had given him his weekends back as he did not need to travel as much.
I have been thinking about that hallway a lot lately, because the industry has just changed in a way that should make stories like this one matter more.
How The Bottleneck Has Moved
For a long time, the bottleneck in software development has been production. We could not write code fast enough. We could not test it fast enough. We could not ship it fast enough.
Now, AI writes roughly 30% of the code at Microsoft. Stripe’s internal agents merge more than a thousand pull requests a week. AI tools give product managers a roughly 40% productivity gain, concentrated in content-heavy work like synthesizing research and drafting documents.
As engineering accelerates, we no longer need to wonder about how product management can speed up.
The new bottleneck becomes the decision-making process to prioritize what is actually worth building, as the cost of being wrong about what to build is higher than ever.
As product discovery coach Teresa Torres explains in a recent interview: “The faster you can ship the wrong thing, the more damage you can do.”
According to Pendo’s product benchmarks, about 80% of software features were rarely or never used all the way back in 2019. This means that most of what software teams were shipping was wrong even before AI made it possible to ship faster.
Why Aggregation Of Data Can’t Match Observation
Product managers spend a significant amount of time aggregating data and anecdotes to prioritize capabilities in the roadmap.
While aggregation will tell you what customers can articulate and what data can surface quickly, it cannot reveal the unmet needs customers don’t know how to express.
Like with my story about waiting in the hallway, the things that kill a product rarely show up in a ticket or a survey.
All too often, a customer hits a problem, finds a workaround and says nothing, because from their perspective, the workaround is fine. Product managers also look at the same bug and make the same call: There is a workaround and the customer is okay, so put the effort into something new instead.
Then those small failures show up again and again in the workflows. The paper cuts accumulate into lost trust, and lost trust is how products quietly lose customers to a competitor.
That is the whole problem with treating aggregation as discovery. What works is often unglamorous. It requires the time and patience to observe users in their natural workflows and understand what they do, not just what they say.
A product manager’s role is not just about looking at multiple tools that aggregate data or identify common themes from customer surveys. It is about investigating the “why” behind the patterns and the context of the user that can only come from observing users firsthand.
If you cannot describe the specific workflow you are improving, you are building on assumptions. Great product decisions come not from aggregation alone, but from combining evidence with empathy, curiosity and deep customer understanding.
The Job That’s Left
The skills that just got more valuable, almost in inverse proportion to the rest of the role, are observation and the judgement to make decisions based on those observations. When execution becomes cheap and fast, the advantage comes from uncovering and solving the right problems.
Early in my career, spending hours with a single customer often felt inefficient. One conversation rarely produced enough evidence to justify a new feature. While it’s true that value often cannot be found in a single insight, it does come from watching how real users work, understanding the context around their decisions and identifying the friction they have with the product.
Today, product managers have no shortage of tools to collect customer feedback and identify patterns at scale. AI can now synthesize those signals even faster and surface trends for broader impact. What it still struggles to capture are the human signals that I first learned to recognize by observing customers in their own environments 16 years ago.
In many ways, the best product managers were the forward deployed engineers of our time. They immersed themselves in customers’ environments, understood their workflows firsthand and translated those observations into products that solved real problems.
I have spent more than two decades building products, and the teams that have always shipped the right things are the ones who never let go of the discipline of being in the room with the user.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
