After two decades working in software development, I’ve never seen product teams pour this much effort, burn through this many AI tokens, and deliver so little tangible, measurable value. This pattern plays out across the entire industry: both large established corporations and early-stage startups are completely reworking their product development workflows to center generative AI today.
Right now, AI handles nearly every step of the modern product pipeline: it drafts project requirements, builds out test frameworks, generates reference datasets, and even ships production-ready code. But so far, there is little clear evidence that all this frantic activity actually results in better end products for users.
There’s an uncomfortable truth at the heart of this AI boom that few people are willing to say out loud: every AI token consumed comes with a real, tangible cost. That cost shows up in cloud compute bills, in rising energy demand, in wasted capital, and most importantly, in drained team attention. We’re burning through non-renewable resources on the backend, and burning out product and engineering teams on the frontend. Carbon emissions from AI workloads are surging, teams are stretched thin to keep up with arbitrary AI targets, and the return on all this investment—for both the planet and the business—is nearly impossible to find.
Forget the Mandate
The common directive to “use AI to deliver customer value” is not a coherent product strategy. It is just a top-down mandate. When I attended the 2026 Fast Company Impact Council Annual Meeting, one CEO put this problem perfectly:
“No one mandated that we use the iPhone. We use it because it works.”
No company, he added, ever earned long-term success by checking a box to top an internal iPhone adoption leaderboard.
Great product decisions have never come from top-down mandates. They come from asking a small set of straightforward, foundational questions:
Do our customers actually need this solution?
Will they be willing to pay for it?
Can we build it well, and will the value it delivers hold up over time?
What makes our team the right group to deliver this?
Will the end result actually delight the people who use it?
And the simplest, most critical question of all—one that most teams skip entirely today: Does this use case actually need AI to work well?
Today, these foundational questions are being shelved entirely to make room for forced “AI-first” features: generic document summaries, unrequested chat boxes bolted onto every product page, and “just-in-time insights” no customer ever asked for. Most of these additions are completely undifferentiated from what every other company is launching, and they’re quickly forgotten by users. They are nothing more than early experiments dressed up as a formal product strategy. They do not create lasting, durable value for the business, and they do not do anything to retain existing customers.
Hard industry data backs up this observation. MIT’s Project NANDA studied the state of AI adoption in global business in 2025 and found that 95% of enterprise generative AI pilots delivered no measurable positive impact on a company’s bottom line. The study’s authors were clear about the root cause of this widespread failure: the problem was rarely the AI model itself. The gap lies between what the tool can do, and how real organizations actually work to serve customer needs every day.
Forget the Mandate
