The quarterly review looked good. AI adoption metrics were up across the board — 73% of employees had completed training, 61% were logging active tool usage, and the initiative was tracking ahead of schedule. Leaders nodded through the slides. Someone mentioned scaling to the remaining business units before year end.
Three floors below that meeting, people were comparing notes in a Slack thread on how to satisfy the usage audit without actually changing how they worked.
This story is not unusual. HBR published research recently showing that executives and employees have significantly different reads on how AI adoption is going inside the same organization. Executives feel positive about the trajectory. Employees feel anxious and substantially less prepared than their leaders assume. The gap has been documented across industries, company sizes, and geographies. And yet every executive who encounters that finding tends to believe it applies to other organizations, not theirs.
The Mechanism Behind the Gap
The gap does not persist because employees are resistant to change or because the technology is poorly designed. It persists because organizations have a deeply embedded reflex for managing information upward. When a VP is visibly enthusiastic about a new initiative, the ambient culture of most large organizations makes one response considerably safer than the other. Showing up with positive news is career-neutral at worst. Surfacing friction, raising doubts, or reporting that the rollout is harder than projected reads as resistance. In some organizations it reads as a political problem.
So people make the rational choice. They complete the training, log some usage, and craft their status updates in the register that organizational culture has quietly rewarded for years. By the time those updates aggregate into a dashboard, they have passed through several rounds of editorial optimism. The executive sees the result and calls it progress.
The failure mode in most AI rollouts is not the technology and it is not the people. It is the information architecture — the organizational tendency to select for good news at every layer between the frontline and the executive floor.
What Executives Are Actually Deciding On
This matters because the decisions downstream of those adoption reports are not small ones. An executive who believes adoption is at 61% will make different resource allocation choices than one who knows that 61% means "completed training and technically logged some usage in the past 30 days." Those decisions compound: expand the rollout, add licenses, announce the initiative as a model for other business units, publicize the win to investors or the board.
Meanwhile, the Slack thread continues.
A close look at the AI initiatives that have genuinely scaled inside complex organizations reveals a consistent pattern. The gap closes faster in places where the information environment makes honest reporting as safe as enthusiastic reporting. That does not happen through more executive communication about the transformation vision or more all-hands reinforcement of the strategic imperative. It happens through structural choices: feedback channels that report upward without attribution, adoption metrics that measure behavioral change rather than tool access, and a cultural signal from senior leadership that surfacing a real problem is more valued than delivering a clean update.
Where to Look Instead
The executives who close the adoption gap fastest are not the ones who communicate most forcefully about the agenda. They are the ones who go looking for the Slack thread.
In practice, that means asking different questions at different levels of the organization. Not "are people using the tools" but "what are they actually using them for, and what happens inside their workflow when the output is wrong." Not "what does the adoption dashboard show" but "who generated that number, and what does adoption mean in the way they counted it." It means having conversations two or three levels below the executive team, where career risk is low enough that something honest might actually surface.
The organizations getting real return on AI investment in 2026 are not the ones with the highest reported adoption rates. They are the ones with the most accurate picture of where adoption is failing and why. That picture does not live in the dashboard. It lives in the places executives have quietly stopped looking — because those places are not in the reporting chain, and nothing about the organizational culture makes it easy to find them.
If your AI initiative is on track according to every metric you are watching, that is worth sitting with for a moment. Who built those metrics, what incentives shaped how the data was collected, and when was the last time you heard something genuinely inconvenient about how it is actually going?
What signals are you actually using to track AI adoption in your organization, and how confident are you in where they come from? I find that question generates more useful conversations than almost anything else right now.
