UpTrajectory Review
The article highlights the challenges faced by product teams in the era of AI, emphasizing that despite the increased use of AI tools, many teams are experiencing burnout and delivering minimal value. The author argues that the rush to integrate AI into product development is often driven by mandates rather than genuine customer needs, leading to a disconnect between effort and outcome.
For small business operators, this piece serves as a cautionary tale about the pitfalls of adopting AI without a clear strategy. It's crucial to evaluate whether AI truly enhances your product or if it's merely a trend-driven addition. The emphasis on AI-first features can distract from fundamental questions about customer needs and product viability. As the article notes, '95% of enterprise generative AI pilots delivered no measurable impact on the bottom line.' This statistic should prompt operators to critically assess their own AI initiatives and ensure they align with tangible business goals.
“95% of enterprise generative AI pilots delivered no measurable impact on the bottom line.” — Fast Company
Takeaway: Evaluate AI initiatives to ensure they meet real customer needs and deliver measurable value.
From the original item — Fast Company:
I have spent 20 years in software development. I have never seen product teams work this hard, burn this many AI tokens, and deliver this little measurable value.
Across the industry, big companies and startups alike are rebuilding how they make products. AI now drafts requirements, writes test frameworks, builds reference data, and ships code. Whether all that activity produces better products is, so far, unclear.
Here is the uncomfortable part. Every token has a cost. Compute, energy, money, and attention. We are burning fuel at the back end and burning people at the front end. The trees are burning and the teams are burning out. The return on both is hard to find.
The instruction to “use AI to deliver customer value” is not a strategy. It is a mandate. At the 2026 Fast Company Impact Council Annual Meeting, one CEO told me, “No one mandated that we use the iPhone. We use it because it works.” No one, the CEO added, was ever rewarded for topping an iPhone leaderboard.
Good product decisions have never come from mandates. They come from simple questions. Do customers need this? Will they pay for it? Can we build it, and is the value durable? What makes us the right team to build it? Will the result delight anyone? And the simplest question of all, the one teams now skip: Does this use case actually need AI?
Those questions are being shelved in favor of AI-first features. Document summaries. Chat boxes. Just-in-time insights nobody asked for. Most are undifferentiated and forgettable. They are early experiments dressed up as strategy. They are not yet durable value, and they are not yet keeping a single customer from leaving.
The data backs this up. MIT’s Project NANDA studied the state of AI in business in 2025 and found that 95% of enterprise generative AI pilots delivered no measurable impact on the bottom line. The authors were clear about the cause. It was rarely the model. It was the gap between the tool and the way real organizations actually work.
That gap is the whole story. Getting good output from AI takes an enormous amount of human intelligence first. Someone has to build the context, define what good looks like, and shape the messy process around it. Human intelligence is scaffolding artificial intelligence, not the other way around. AI fails without it.
The bar keeps rising. In 2025, prompts became workflows. In 2026, workflows became agents. Customers no longer want a clever answer. They want an agent to execute a validated workflow correctly, every time. That is a far harder problem. It forces businesses to write down the tribal knowledge that has never been written down, to map how work really happens, and to define the standard an agent must meet. Most of this is slow, unglamorous human work. Humans have to agree with humans before agents can be trusted to act.
This is a new and heavy burden. It lands on product teams first, and on every business hoping AI will solve its problems. It demands proofs of concept, patience, and a tolerance for unpredictable surprises that few companies actually have. So teams absorb the gap. They work nights. They burn tokens. They ship features they do not believe in. That is what burnout looks like.
There is a way out, and it is not new. Go back to basics. Before the next AI sprint, make the team answer five plain questions. Who are we building for? What is the job to be done? Do we know what good looks like? What is the best way to solve this, with or without AI? How fast can we prototype and test it?
If the honest answer to “does this need AI?” is no, that is a win, not a failure. It saves tokens. It saves money. It saves your best people. The trees and the teams will thank you.
Anubhav Rohatgi is senior director of products at Adobe Inc.