Trust is one of the most important things in an employment relationship. It is also one of the easiest to lose — not through a single dramatic event, but through a slow accumulation of things left unsaid.
Right now, in most organisations, what is (not) happenning in terms of AI, is doing exactly that.
Both sides are hiding something
Look at what is actually happening on the ground. Microsoft and LinkedIn’s Work Trend Index found that around three quarters of knowledge workers already use AI at work — but more than half of them are reluctant to admit using it for their most important tasks, because they worry it makes them look replaceable.
Read that again. Your best people have found ways to work faster and better, and they are hiding it from you. Not because they are dishonest, but because they have done the maths: if I show how much of my job AI can do, I may have just written my own redundancy notice.
Meanwhile, HR is running its own concealment programme. The official line in most companies is some version of “AI is not going to take your job.” It is meant to be reassuring. It is not. Employees are not naive — they read the same news, they use the same tools, they can see perfectly well which parts of their work are changing. When the official message contradicts their daily experience, they don’t conclude that everything is fine. They conclude that nobody is telling them the truth.
So both sides are performing calm while quietly preparing for a future they refuse to discuss with each other. That is what eroding trust looks like from the inside.
And here is the part that should worry leadership: recent research published in Harvard Business Review found that whether employees disclose how they use AI is predicted far more strongly by organisational trust and psychological safety than by any formal AI policy or approved toolset. You cannot fix this with a policy document. It is a trust problem, and it has a trust-shaped solution.
What honest actually looks like: Klarna
At a recent conference, Andrea Dintser made a point about Klarna that stayed with me. Everyone knows the headline version of that story: Klarna went AI-first in customer service, shrank its workforce dramatically, then publicly admitted it had cut too deep and started rebuilding human capacity.
Most commentary treats this as a cautionary tale about AI. Andrea’s point was sharper: the remarkable part is not the overreach — plenty of companies overreach. The remarkable part is that Klarna communicated from point zero. Openly, publicly, including the uncomfortable parts. They said: this is our plan, this is what is happenning, this is what worked, this is what didn’t, this is how we are fine-tuning. That is rare. And it is precisely that willingness to keep communicating and keep adjusting — rather than defending a position — that turns a mistake into a credible course correction instead of a credibility collapse.
What proactive looks like: Orange
Orange went further, and earlier. Three years ago, they assessed which roles in their organisation were most likely to be displaced by automation. Then they did the thing most companies find unthinkable: they sat down with the people actually holding those jobs — before anything was decided over their heads — and agreed on reskilling paths and a new deal for the employees affected.
They have presented on this work openly, and they have kept building on it: their current employment and career-path agreement, signed unanimously with the unions, continues exactly this logic — anticipating which roles will change and opening voluntary retraining routes toward the professions that are growing.
Notice what Orange did not do. They did not promise that nothing would change. They did not hide the assessment results from the people affected. They treated employees as adults who can handle the truth — and who, given a real path forward, will usually take it.
Trust is not protected by reassurance. It is protected by a plan.
This is the conclusion I keep arriving at. The instinct in HR is to protect trust by softening the message. But vague reassurance is not kindness — it is a deferral, and everyone can feel it.
What actually protects trust is a credible, honest answer to the questions employees are already asking silently: Which roles will change? By how much? What are the realistic paths from here to there? What will the company invest in getting me there?
You cannot answer those questions with sentiment. You can only answer them with a map — a strategic workforce plan that shows, role by role, what automation and augmentation actually mean for your organisation, who can move where, what it costs, and in what order. A plan solid enough to put on the table in front of the very people it concerns. That is the standard Orange set, and it is the standard the Klarna story teaches by contrast.
AI is not the thing eroding trust in our workplaces. Silence is. And unlike the technology, the silence is entirely within our control.
Plan the transition with the people it concerns, on real numbers.