There is a particular kind of leadership failure happening inside organizations right now that almost nobody is naming clearly, even though it is producing some of the most consequential workforce anxiety of the decade.

It is not a failure of strategy in the conventional sense, and it is not a failure of technology adoption, and it is not even really a failure of communication, although it gets diagnosed as one most of the time. It is a failure to articulate a deal, and the consequences of that failure are showing up in trust scores, in disengagement, in quiet attrition, and in productivity gains that mysteriously fail to materialize even after the AI tools have been fully rolled out and the training has been fully delivered.

The pattern is familiar to anyone paying attention. A leader stands in front of their team, or sends an all-hands memo, or drops a line in the company chat about how everyone needs to be using AI more, faster, and smarter, and the room nods politely, and then the team goes back to their desks and starts quietly updating their resumes. Not because anyone hates the technology, and not because the staff is resistant to change, but because nobody has actually told them what the deal is, and in the absence of a clear deal, they are doing the math themselves and arriving at the worst possible answer.

Call it evangelism without a compact, and recognize it as one of the most pressing leadership failures of the moment. The data is unambiguous on the scale of the problem.

The Mercer Global Talent Trends 2026 survey of twelve thousand workers and business leaders found that forty percent of employees now fear losing their job to AI, a sharp jump from twenty-eight percent just two years ago. The KPMG American Worker Survey from late 2025 found that fear of job displacement due to AI has nearly doubled from the previous year, with fifty-two percent of workers now expressing concern that AI could eventually replace their jobs, even as nearly nine in ten workers reported using AI at least weekly. And a national survey from Jobs for the Future, released in March, found that workers are now more likely to say AI is a net-negative than a net-positive when it comes to finding jobs and building wealth, a sentiment that flipped in a single year. Just over one-third of workers say employers are providing the training, guidance, or opportunities they need to use AI in their jobs, which is a drop of almost ten percentage points from the previous year.

The workforce is anxious, yet the workforce is using these tools anyway. The workforce is not getting much in the way of clarity from the people at the top who are supposed to be leading them through this. That is the gap worth talking about, because many leaders are mistaking enthusiasm for strategy and confusing a mandate with a plan.

The version of the failure that seems to be doing the most quiet damage is the leader who tells their staff to run something through AI to make sense of it, before the staff have even tried to make sense of it themselves, and sometimes before the leader has read the thing at all. There is a particular flavor of this that is genuinely troubling, which is when a leader is essentially using AI as a stand-in for trusting their own people, asking the model to summarize a memo their direct report wrote, or to evaluate a strategy their team proposed, or to take a first pass at a decision that really ought to be informed by the humans who have been thinking about it for weeks. The signal that sends to the team is not subtle, and it is not the signal the leader thinks they are sending. The signal is that the leader trusts the tool more than the person they are paying, and once a team feels that, it cannot easily be walked back.

The research on this is sobering. A study published in the International Journal of Business Communication examined perceptions of low, medium, and high AI use in workplace writing, and found that when supervisors heavily used AI for routine communication tasks, respondents began to question the sender's authorship, confidence, caring, sincerity, and ability.

Which is to say, when the boss leans too hard on AI, the team stops trusting the boss. The Center for Creative Leadership, in their work on trust and AI transformation, makes a similar point, arguing that AI is destabilizing the foundations of workplace trust at the exact moment organizations need unprecedented levels of trust to adopt it successfully. When leaders use AI to draft communications, the question employees are silently asking is whether to experience it as efficiency or as a diminishment of respect, and a lot of leaders are not even aware that question is being asked.

The reflexive response to all of this, from leaders who are well-meaning and who genuinely do not want to be the cause of staff anxiety, is to offer reassurance. AI is just a tool. It is not going to replace you. It is going to make your job easier and your life better and your work more interesting. The problem is that this reassurance almost never works, and it does not work for a reason that should be obvious to anyone who has been paying attention to how teams actually process information. In the same week that a leader is saying AI is just a tool, that same leader is also saying, in earnings calls and strategy memos and board presentations, that the organization needs to be leaner and faster and more efficient, that headcount is under review, that productivity gains are non-negotiable, that the future belongs to companies that can do more with less. The team can hear both of these messages at the same time, and the team is not stupid, and so the team correctly identifies which message is the actual plan and which one is the public relations layer on top of it.

The trust gap research reinforces exactly this dynamic. A piece in People Managing People made the point sharply, arguing that AI adoption outcomes across sectors correlate almost perfectly with pre-existing trust levels between leadership and staff. Where employees felt they had been communicated with honestly during previous periods of change, they are more optimistic and more willing to give AI a chance. Where the opposite happened, where decisions were made in a vacuum without communication, AI becomes what one leadership coach quoted in the piece called a lightning rod of conflict. That is when employees start asking whether the rollout is really about productivity or about cutting headcount, and they start feeling expendable, and once that feeling sets in, no amount of reassurance is going to fix it because the reassurance is not the problem.

What is required instead, and what most leaders are not doing, is a compact. Meaning a clear, repeated, specific articulation of what the deal actually is. Not a memo, not a town hall, not a slide in a quarterly all-hands, but an ongoing and explicit conversation about what AI is for in this organization, what humans are for, what the trade is, and what the commitment is to the people who are on the other side of that trade. The reason this matters so much, and the reason vague evangelism is so damaging, is that staff are going to fill in the silence themselves, and the silence almost never gets filled in with optimism.

At Field Nine Group, the parent company of Adriana Lacy Consulting, the way the compact has been articulated, and the language that gets repeated internally often enough that the team is probably tired of hearing it, is that AI is a layer that allows the company to move fast on the things that should not require human time, and the purpose of that speed is to give the team more time, not less, to spend on the work that actually differentiates the firm. I tell my team constantly that AI handles task assignment and forecasting and first-pass research and meeting summaries and the kinds of operational drag that historically eat the calendar of a senior strategist. The team handles strategy, design judgment, editorial taste, client relationships, the human-centric work that nobody is paying Field Nine for AI to do. AI is the layer, and the team is the work, and the formulation functions less as a slogan than as a cultural anchor that the team can hold leadership to.

That is the part most leaders miss when they think about what a compact is for. It is not a marketing exercise, and it is not a values statement. The point of putting the deal in writing, and saying it out loud often enough that it becomes part of the shared vocabulary, is to create something the team can use to push back when leadership starts to drift. If a leader with that kind of stated compact were to suddenly start asking the team to let AI take the first pass on a client strategy, the team would have grounds to point out that this is not what the compact said AI was for, and leadership would either have to explain why the line has moved or admit that the plot has been lost. A vague mandate does not give a team that kind of footing. A vague mandate is just a permission slip for the leader to keep moving the line and for the staff to keep absorbing the consequences.

A recent piece in CEOWORLD by David Grossman, who has spent twenty-five years advising CEOs through transformation, captures the underlying issue almost perfectly. He makes the point that the question every employee is silently asking during an AI rollout is whether AI is their partner or their replacement, and that if their leader has not answered that question, both with words and with consistent actions, people will fill in the worst answer themselves and behave accordingly. He also cites BCG research showing that at leading future-built firms, fifty percent of employees are expected to be upskilled in AI by year-end, while at lagging firms that number is twenty percent, and the gap is not about budget. The gap is about whether employees believe there is a future for them on the other side of the upskilling. That is the compact question, exactly. Do the people inside the organization believe leadership has made room for them in the version of the company being built, or do they believe leadership is building something that does not require them.

A lot of leaders, if they were being honest, have not actually decided the answer to that question yet, and that is the real source of the vagueness. It is not a communications problem. It is a strategy problem dressed up as a communications problem. The compact requires leaders to make decisions they have been avoiding, to commit to specific use cases instead of broad enthusiasm, to be honest about what they are trying to optimize and what they are willing to protect, and to put that in writing in a way that their team can hold them to. That is much harder than evangelizing, which is why so few leaders are doing it, and which is also why the staff anxiety is not going to be solved by another all-hands.

A piece in Harvard Business Review from earlier this year put a finer point on the diagnosis than most of the trade press has managed. The authors noted that many leaders treat AI integration issues as technology problems to be solved with better tools or training, rather than team effectiveness issues that require many of the principles already known to work for human collaboration. That framing is exactly right, and it extends further than the HBR piece itself goes. The reason an AI rollout is creating anxiety is not that leadership picked the wrong tool, and it is not that the training is insufficient, and it is not that staff is resistant to change. The reason an AI rollout is creating anxiety is that leadership has not yet done the work of deciding what the organization is actually for in a world where AI exists, and the team can sense the absence of that decision, and they are responding to it the way humans always respond to ambiguity from people in power, which is to assume the worst and protect themselves.

Anyone can be excited about AI right now. The market is excited, the trade press is excited, the board is excited, and there is no professional cost to riding that wave. The professional cost comes from the staff who quietly disengage, the institutional knowledge that walks out the door, the productivity gains that fail to materialize because nobody actually trusts the system enough to use it well, and the culture that hardens against future change because this particular change was handled so poorly. Those costs are not visible in the same quarter as the AI investment, which is part of why leaders keep getting away with skipping the compact, but they show up eventually, and by the time they show up the damage is done.

For any leader reading this and recognizing the pattern, the move is not a new policy or a new memo or a new training program. The move is to sit down, alone or with one or two trusted operators, and write out in plain language what AI is for in the organization, what the people are for, what the trade is, and what leadership is committing to do for the humans who are on the other side of the change. Then say it out loud to the team, and keep saying it, and let the team push back on it, and refine it, and treat it as a living agreement rather than a one-time announcement. That is what a compact looks like. It is not glamorous, and it will not get a leader on a panel, but it is the difference between an AI rollout that builds an organization and one that quietly hollows it out.

This is going to be one of the defining leadership questions of the next eighteen months, and most organizations have not figured it out yet. But the answer almost certainly starts with a compact, and the leaders who get this right in the near term are going to have a meaningful advantage over the ones who keep evangelizing without one.