Ask in a retro how the AI rollout is going and you'll get "fine, mostly." Ask via survey and you'll get 6.4 out of 10. Ask the team lead one-on-one and you'll get a diplomatic version of the truth. All three answers are sincerely meant. None of them are honest.
Not because people lie, but because everyone filters the moment someone with influence over their career is in the room. That's not a flaw; it's social intelligence. And that very filtering makes organizations blind at exactly the moment they most need to see - when a technology changes the rules and nobody is sure what it means for their own job.
Three Types of Silence
Every AI transformation produces three groups that don't say what they think. Their reasons have nothing in common.
One group is struggling. The tool overwhelms them, but they can't admit it. In an environment where AI proficiency has been declared a core competency, "I can't keep up" is a career risk.
So they smile during training, nod in the retro, and google the basics at home in the evening. BCG's AI at Work Report 2025{target=_blank} puts numbers to the gap: only about a third of employees consider AI training sufficient. 18 percent of regular users never received any training at all. The tool is there, the expectation is there; the enablement is not.
Another group could be much further ahead. They've built their own workflows, developed prompts, pushed the tool beyond what anyone intended. They just don't show it.
Getting too far ahead creates friction - with the manager who hasn't thought that far, with colleagues who then feel even more inadequate. So the best hold back, and the organization never learns what's possible.
The most dangerous group is invisible. These people are neither overwhelmed nor enthusiastic. They simply don't understand what's supposed to change for them.
They attended the training, saw the slides, got access - and went right back to doing things the old way. Not out of resistance, but out of confusion. Their silence looks like agreement from the outside, so nobody addresses it.
Three groups, one shared problem: the topics that most urgently need discussion are exactly the ones nobody talks about. And the harder management pushes for adoption, the quieter it gets. The culture dissonance that builds in the process often stays invisible.
The Adoption Lie
73 percent active usage. Training completed. Pilot phase officially successful. In the steering committee, the rollout is presented as a win. That's what the dashboard says.
What the dashboard doesn't say: three out of eight only open the tool when someone is watching. The person who actually works with it doesn't share her prompts.
Everyone nodded in the retro. In reality, almost nothing has changed about day-to-day work - only the statistics look better.
Gartner has a name for this: Regrettable Retention. People are physically present, show up in every metric as active users, and have mentally moved on from the transformation long ago. A pattern that also surfaces in employee interviews: the official numbers tell a different story than the real reasons.
The adoption rate can sit at 90 percent while actual value creation stagnates at zero. The gap between "the tool is installed" and "the tool changes how we work" is enormous; no dashboard in the world captures it.
The problem isn't that companies measure the wrong metrics. It's that the right information only emerges in conversations that aren't happening.
What Changes When Hierarchy Leaves the Room
"How has your workday changed since the AI rollout?"
Ask that in a retro and you'll get diplomacy. In a setting where no supervisor is listening, no transcript ends up in a personnel file, and no face can be lost, you get something different.
You get the shadow workflows: "I do the task manually first and then feed the result into the tool so it looks like I used it." Or the hidden enthusiasm: "I taught myself how to automate reports in the evenings, but I don't mention it in meetings because everyone will think I'm showing off." And the overwhelm that has no name: "I don't understand what the tool does better than my current approach. And I don't dare ask."
These aren't footnotes. These are the insights you can actually make decisions on - about training formats, about team dynamics, about whether the rollout is actually working or just looking that way.
For answers like these to surface, three things are needed: anonymity, depth, and the right timing. Anonymity alone isn't enough; an anonymous questionnaire with checkboxes produces "more training would be nice" without clarifying what kind of training, for whom, and why.
What's missing is the follow-up question. The moment someone asks "What do you mean by more training?" and the real answer comes: "I don't need training. I need someone who shows me it's okay to make mistakes."
An interview that listens and follows up, conducted asynchronously, without scheduling pressure, without social context - that changes what people are willing to say. Not because it manipulates them, but because it creates conditions where honesty isn't a risk.
When Listening Makes the Biggest Difference
Four weeks after go-live is the most critical moment. The initial curiosity has faded, the training effects are wearing off, and routine has taken over.
This is when the workarounds appear, when enthusiasm tips into overwhelm, when it becomes visible which workflows actually work and which only exist on paper. If you don't listen at this point, you miss the window where course corrections still make a difference. Continuous Discovery turns this kind of listening into a permanent part of the process.
Six months later comes the harder question: has the way people work actually changed, or are they just feeding the dashboard? Have people developed new routines, or have they preserved old ones and dressed them up with AI cosmetics? If you're asking for the first time here, you're asking too late.
But even before launch, there's a conversation most teams skip. People carry expectations and fears that never get voiced.
"Will my job become obsolete?" coexists with "Finally, I won't have to do that reporting garbage by hand anymore." Knowing that tension before the rollout begins lets you tailor the introduction accordingly - instead of broadcasting into a room that has already gone quiet.
The Question That Remains
The most successful AI rollouts won't be the fastest ones. They'll be the ones where someone listened systematically - before the workarounds hardened, before the quiet ones concluded that nobody cares about their perspective, before the silence became so normal that nobody noticed it anymore.
The question isn't whether your team has problems with the AI rollout. The question is whether you're hearing about them.
Sources
- BCG: AI at Work: Friends and Foes - Report 2025- Gartner: Top 9 Future of Work Trends for CHROs in 2026
Try It Yourself
What does an interview feel like that listens and follows up instead of ticking checkboxes? Try it yourself.
