When Workers Trust Neither The Boss Nor The Algorithm

Walk into almost any MBA classroom this term and you will see something quietly strange. A student uses AI to prepare the case, to stress-test the professor’s framework, to draft the market analysis, to rehearse the final pitch. Ask that same student whether they trust AI, and many will say no. Not really. They worry about hallucinations, about bias, about who is accountable when the model is confidently wrong.

They use it anyway. This is the paradox we have not yet had the courage to confront, and it does not stay in the classroom. That student is tomorrow’s manager. The reflex they are building now, leaning hard on a system they do not believe in, is the same reflex reshaping every organization they will walk into. We keep treating this as a question about technology. It is a question about trust. AI is simply the thing that finally exposed it.

THE NUMBERS ARE NOT SUBTLE

The largest global study of its kind makes the gap visible. In the 2025 KPMG and University of Melbourne research, led by Professor Nicole Gillespie and Dr. Steve Lockey, more than 48,000 people across 47 countries were surveyed. Two thirds, 66%, said they intentionally use AI on a regular basis. Fewer than half, 46%, said they are willing to trust it. Adoption and trust, which for most of modern history moved together, have come apart.

Now hold that against what is happening to human authority. The 2025 Edelman Trust Barometer, covering more than 33,000 people across 28 countries, found that “my employer” remains the most trusted institution in society, at 71%. Reassuring, until you separate the institution from the individuals who run it. Faith in business leaders has fallen sharply, down more than a fifth since 2021, and 7 in 10 people now believe leaders routinely tell them things that are not true. Trust in the company holds. Trust in the boss is thinning.

Put those two findings side by side and you have the defining fact of management today. Employees are increasingly asking an AI for advice they would no longer ask their manager. Not because they trust the machine. They have told us, plainly, that they do not. They ask it because it is fast, available, and does not judge them, and because confidence in the human alternative has worn through.

That is reliance without belief. It is the most important pattern in the entire AI conversation, and almost nobody is leading to it, or teaching to it.

WHY THIS IS NOT AN AI STORY 

Here is the reframe, and it comes from how value behaves, not from how technology behaves.

When the cost of producing something collapses toward zero, that thing stops being scarce, and what is no longer scarce is no longer valued. AI has collapsed the cost of analysis. It drafts the recommendation, finds the pattern, synthesizes the report, in seconds, at the margin, for free. So the analysis is no longer where the value sits. The value migrates to what stays rare: the judgment to decide which answer to act on, and the credibility that makes other people willing to act alongside you.

This inverts a century of both management and business education. We have trained leaders to be the person with the best answer. The machine now produces a serviceable answer on demand. The leader’s remaining advantage is not the answer. It is being the person whose judgment others will follow when the answer is already on the table and nobody is quite sure they believe it.

Recall the distinction that still does the most work in this debate. Authority is the right to be obeyed. Trust is the willingness to be vulnerable to someone else’s judgment. AI has accumulated authority at extraordinary speed, because it offers exactly what organizations prize: scale, consistency, and the appearance of objectivity. What it has not accumulated is trust. And so we arrive at a genuinely new organizational condition, one that deserves a name: decisions that are followed without being believed.

THE BOSS IS BECOMING AN INTERFACE

For most of the modern era, a manager’s job was to add judgment to information. The information was scarce; the judgment was the value. AI has reversed both. It supplies the information and a judgment-shaped output to go with it, instantly and at almost no cost. A manager who simply relays what the model recommends has quietly become something smaller: a human interface for an algorithm, an escalation path with a pulse.

This is the real exposure for leadership, and it is why “the employer is still trusted” should comfort no one. Authority that is obeyed but not believed is brittle. It holds until the first decision that genuinely tests it, and then it does not. If employees trust neither the boss nor the machine, leadership itself is the casualty. That is the sentence we would hang on the wall of every executive program.

Which is why trust stops being a soft virtue and becomes a hard asset. When the analysis is free, trust is what converts an answer into committed action, the scarce input that decides whether a recommendation moves an organization or just sits in a deck. Leaders who can generate it will compound advantage. Leaders who cannot will preside over teams that nod, comply, and quietly route around them to the chatbot.

THE TRUST CRISIS IS BI-DIRECTIONAL

Everyone frets that we do not trust AI enough. The harder, and in our view more dangerous, truth is that we already trust it too easily wherever it is convenient. The same global study found that roughly 57% of people accept AI output without checking whether it is accurate, and around 59% admit it has led them into mistakes at work. A 2025 study by researchers at Microsoft and Carnegie Mellon identified the mechanism: the more confidence knowledge workers place in an AI, the less critical thinking they apply to what it gives them.

So the trust crisis runs in both directions at once. Too little trust in the human who is supposed to be accountable. Too much trust in the system that cannot be. And the casualty caught between them is judgment itself, in the team and in the leader.

We do not say this from the stands. A few weeks ago, one of us spent a day with around forty academic leaders from across Europe, convened to discuss precisely this. The room kept returning to the same words the sector has been naming every year since 2023: speed, agility, trust, employability, the erosion of academic integrity. Naming them has become a reassuring ritual. The diagnosis is impeccable. The courage to act on it is what keeps going missing.

WE ARE VERY GOOD AT TEACHING AUTHORITY

Business schools have become excellent at teaching authority. We teach strategy frameworks, financial models, analytics, operations, and now prompt engineering. These matter, and they will keep mattering.

What we barely teach is trust, because we still treat it as a personality trait. Some people have it, some do not, and there is supposedly little to be done. That is simply false. Trust is a competency. It is built, in public, under pressure, and it can be trained like any other. Leaders earn it when they explain the reasoning behind a hard call instead of hiding behind the result, when they name the uncertainty instead of dressing it up as confidence, and when they take the consequences rather than routing them elsewhere.

Those are exactly the things a model cannot do for you. A model cannot stand in front of a room and be accountable.

A CURRICULUM OF TRUST

So here is the design principle, the one idea we would ask every dean to sit with. The decisive line does not run between the human and the machine. It runs definitely inside how we use the machine, between the AI that spares us the effort of thinking and the AI that forces us to think better. Build the curriculum on the second one. We do not teach judgment by lecturing about it. We teach it by making students decide, defend, fail, and own the result. The thinking is born from the doing.

Concretely, that means at least five things faculty and program directors can build now:

1. Decision defense. The student presents an AI-assisted recommendation and must say out loud what they accepted from the model, what they rejected, and what they are personally accountable for. Prompting is the easy half. Owning the call is the half we grade.
2. Authentic assessment. The oral, the defended project, the viva. Formats a model cannot sit in the student’s place, and the most direct way to restore trust in what a degree actually certifies.
3. AI as a friction partner, not a prosthesis. Assignments where the model is required to argue against the student, surface the objection they missed, and defend the rival frame, while the student still has to make and own the decision. As a bonus, it is one of the few real antidotes to the sameness that AI-generated work drifts toward.
4. Accountability under uncertainty. Simulations with real stakes and no clean answer, where the choice is made in public and the consequences are carried, not hypothetical.
5. The right to refuse. Structured practice in disagreeing with an algorithmic recommendation, and in saying no to it with reasons. A leader who cannot overrule the system is merely its interface.

None of this is theoretical, and that is the point we would press on anyone who says it cannot be done at scale. At NEOMA, where one of us leads the school’s digital strategy, roughly 12,000 people were trained, ninety percent of the faculty and half the staff, in three years. Not because a committee issued a philosophy. Because a clear vision was taken, and following decisions were executed, and measured. Philosophy, we find, arrives while you walk.

THE CASUALTY, IF WE GET THIS WRONG

The stakes are about to rise, not fall. Employees and students alike already orchestrate AI agents that do not merely answer but act: they write, they trigger workflows, they execute. When a person delegates a decision to a system that then goes and acts in the world, the question of who remains accountable does not get easier. It gets sharper, for the manager and for the school that trained them.

Which returns us to the only durable answer. The leaders who matter in the AI era will not be the ones who use the tools best. Plenty of people will use the tools well. They will be the ones whose judgment stays credible after the algorithm has already spoken, the ones their people are still willing to follow into a decision nobody fully understands.

Authority can be assigned. Trust must be earned. For business schools, teaching that, and not teaching the next prompt, may be the most important work of the decade ahead. For the leaders they send out, it may be the only thing the machine cannot do in their place.

---

Benjamin Stevenin is the former Director of Business School Solutions and Partnerships at Times Higher Education. Alain Goudey is a marketing professor and Associate Dean for Digital at NEOMA Business School in France. 

© Copyright 2026 Poets & Quants. All rights reserved. This article may not be republished, rewritten or otherwise distributed without written permission. To reprint or license this article or any content from Poets & Quants, please submit your request HERE.