Is AI quietly eroding the trust inside your team?

Someone sent you a flawless document last week. Clean structure, confident prose, not a comma out of place. And your first thought was not well done. Your first thought, the one you did not say out loud, was: did you actually write this, or did the machine? You sat with it for a second longer than you would like to admit. Then you said nothing, replied "great, thanks," and moved on. But something had shifted, and you felt it.

That small, unspoken question is the new thing AI has brought into your team. AI trust in teams is not really a question about the technology at all. It is a question about whether you still believe the work in front of you came from the person whose name is on it, and whether they believe the same about you. Trust is the operating frequency a team runs on. When that frequency drops, everything downstream slows: decisions, candour, the willingness to put a rough idea on the table. And right now, quietly, it is dropping.

Why does a polished document make you suspicious now?

Because the suspicion is well founded, and it runs in both directions. A 2025 study from Duke's Fuqua School of Business, across nearly 4,500 participants in four experiments, found a measurable "social evaluation penalty for using AI": people who used AI were consistently rated as lazier, less competent, and less diligent than those who got help from a human or from an unspecified source. Read that again. Your colleague who suspected the machine wrote your report is not being paranoid. The judgement is real, it is widespread, and people know it exists.

Which is exactly why the document arrived with no mention of how it was made. The same Duke researchers found that because people anticipate this penalty, they become far less likely to disclose their AI use to managers and colleagues. So you have two truths sitting in the same room: the suspicion is reasonable, and the silence is rational. Each one feeds the other. You wonder if the machine wrote it; they say nothing because they know you will wonder. That loop is the symptom, and you can see it almost everywhere now.

How widespread is the hiding, really?

Wider than most leaders assume, because the people doing it are the ones least likely to tell you. In KPMG and the University of Melbourne's 2025 global study of more than 48,000 people across 47 countries, 57% of employees said they hide their use of AI and present AI-generated work as their own. More than half. The same study found only 46% of people globally are willing to trust AI systems at all, and 66% rely on its output without checking it for accuracy. So your team is using a tool it does not fully trust, to produce work it then conceals, in an environment where being found out carries a competence penalty. That is a lot of weight for a team to carry in silence.

Slack's 2024 Workforce Index, surveying 17,372 desk workers across 15 countries, puts numbers on the feeling itself: 48% would feel uncomfortable admitting to their manager that they used AI for common tasks. The top reasons were that using AI feels like cheating (47%), the fear of being seen as less competent (46%), and the fear of being seen as lazy (46%). Laserfiche's 2025 survey of 1,000 US workers found the same pattern at home: 49% who use AI keep it to themselves, and 15% deliberately avoid telling their manager. This is not a tooling gap. It is a trust gap, and it has moved into the spaces between your people.

What kind of fix actually rebuilds trust here?

Start by naming what this is. The instinct is to reach for a policy: an acceptable-use document, a disclosure checkbox, a line in the handbook. Policy has its place. But the trust gap is wider than any rulebook can close on its own, and it runs straight down the org chart. Checkr's 2026 Manager-Employee AI Divide report (a survey of 3,000 employed Americans, split evenly between managers and staff) found 70% of managers trust AI-driven tools, against just 27% of employees, and that 70% of managers and 58% of employees already believe AI is enabling a new form of workplace fraud. Read those together and the picture sharpens: the people most comfortable with the tool sit at the top, and the people most anxious about being judged for it sit below. The bottleneck is no longer the technology. It sits in the human layer, in whether people feel safe to be honest about how the work gets made.

The category of intervention that works is making AI use visible and ordinary, led from the top. The most striking finding in the Duke study points straight at this: the social evaluation penalty diminished when the evaluating manager used AI themselves. When the person doing the judging is openly in the same boat, the judgement loses its charge. And there is no shortage of leaders who could model it. UpGuard's 2025 shadow-AI study (shadow AI being unapproved AI tools used outside company sanction) of 1,500 workers across seven countries found more than 80% use unapproved AI at work, and that executives report the highest rate of regular use of anyone. The honesty you want from your team is honesty you are already practising in private. That is the lever: not a rule that drives the behaviour further underground, but a leader who names what they already do, so disclosure stops costing anyone their standing.

1. Go first, out loud. Say where you used AI this week and where you chose to think it through yourself. When the leader discloses, the penalty fades for everyone beneath them; the Duke research shows this directly.
2. Separate the tool from the judgement. Make it explicit that quality and ownership are what you value, and that AI is one more way to reach them. This loosens the "cheating" frame that 47% of workers still carry.
3. Build the conversation, then the document. Agree how your team will note AI assistance so it becomes ordinary rather than confessional. The aim is shared language, so honesty costs nothing.
4. Restore your own coherence first. A team reads its leader's state before it reads any memo. Coherence here means head-heart alignment, a settled internal state rather than a performed calm. The peer-reviewed autonomic research is suggestive: higher vagally-mediated heart rate variability, the beat-to-beat variation that tracks a regulated nervous system, lines up with better decision-making under uncertainty and stronger executive control (Forte et al., 2021; 2022). A calm, coherent leader makes a calmer, more candid room. Treat this as exploratory, and treat it as worth your attention.

It helps to see where this sits in the larger arc. Phase One, the Age of Effort: work hard, get a little more, linear growth. Phase Two, the Age of Scale: build once, sell to millions, exponential growth. Phase Three, the Age of Acceleration: output decoupled from human effort almost entirely, the phase AI unlocks. Phase One was muscle. Phase Two was machine. Phase Three is mind, and the work of Phase Three is mostly human: rebuilding the conditions in which people can be honest with each other about a tool they all use and few will name. The leaders who win the next decade are the ones who upgrade themselves first, then make it safe for the team to follow. The machine did not introduce the suspicion. The silence we let grow around the machine did. And silence is something a leader can choose to break.

About the author

Thomas Green

British technology futurist, AI keynote speaker and advisor. Thirty years across enterprise technology and AI strategy, helping leaders navigate the future of work. The futurist who died.