Use AI, Move Faster, Cut Cost. Then What?

Your AI Mandate Measures the Wrong Thing

I was speaking at a product conference last week. I talked to so many people who were in attendance from large and small organizations, and one thing kept coming up: everyone seems to be getting the same directive from the top. Use AI. Move faster. Find the efficiencies. Bring the cost down. It’s in the all-hands decks, in the quarterly goals, in the small nudges inside performance conversations: are you using the tools, what did you automate this quarter, how much time did you save? I have one of these mandates myself, so I am not watching this from the outside. I feel the pull of it like everyone else.

And here is what the mandate has us all doing. We are burning tokens to see what sticks. We throw AI at the first pass of the PRD, the research synthesis, the email drafts and summaries, the presentation decks, and everything comes back looking good. The activity and the output are real and make us feel like we are making progress. Is any of it ‘good enough’ to endure the test of whether we are solving a real user problem? I am not sure we are focused on measuring that.

The number nobody wants to sit with

A few months ago, MIT’s Project NANDA 2025 report, The GenAI Divide: State of AI in Business, looked at enterprise generative AI and found that 95% of organizations are seeing no measurable return on what they have spent, and that is after billions of dollars poured into these initiatives, with nothing measurable to show for it.

The lead author, Aditya Challapally, stated that the failures had very little to do with model quality and almost everything to do with how organizations were trying to use them. He called it a ‘learning gap’, the gap between what the tool can do and the workflow it lives in. The pilots don’t scale well because they have to integrate with existing enterprise systems.

The mandate measures the easy things

The way I see it, the mandate measures whatever is easy to count. Are people using AI? Are we delivering more user story points? Has our velocity increased? All of that is activity. None of it tells us whether we delivered value with all this activity. “It is faster and cheaper” is a different sentence from “it is good,” and a mandate built on speed and cost will happily reward the first while ignoring the second.

That gap is the whole story behind the 95%. We measure the burning of tokens that shows activity, but never set the bar on whether the output was equal to the outcome (hint: It isn’t).

The misalignment was always there

There is a second cost that’s not on the velocity dashboard. When every team burns tokens against the same vague instruction, we do not get more aligned. We get more output, pointed in different directions. Each team builds its own artifacts with its own copilot and never reconciles them with the other teams. So silos deepen. Everyone is optimizing for what is getting rewarded, and right now what’s getting rewarded is more AI use.

Ovetta Sampson said, “Generative AI reconfigures the past and calls it the future.” That is exactly what an AI mandate without clarity does. We take the dysfunctional way we already work, automate it, and call it transformation. The misalignment was always there. AI just lets us produce it at a scale and at a speed we could not reach before.

What worries me most is the slide from outcomes back to outputs. We spent years in product trying to measure whether the work mattered, not just whether the work shipped. A mandate to use AI, measured by usage and speed, pulls us straight back to counting outputs.

So what are the 5% doing that the rest of us are not? That’s next.