What You’ll Find This Week

The standard explanation for why AI is tearing companies apart is that the rollout went wrong. The company moved too fast. Employees got anxious. Leaders underestimated the change management problem. The technology outran the organization.

Maybe. But I think there’s a more useful read.

If AI is creating power struggles inside the company, it may be because AI has finally reached something that matters. Not email drafts. Not meeting summaries. Not a chatbot bolted onto an old process. The real stuff: decisions, approvals, budgets, staffing, accountability, and who gets to decide what counts as true.

That’s where AI stops being a productivity tool and starts becoming an operating-model problem.

This week, I’m looking at a WRITER survey finding that 54% of C-suite executives say AI is tearing their company apart, and why the companies feeling that tension may actually be closer to real transformation than the ones where everything feels calm.

Here’s what you’ll find:

Don’t Miss Our Latest Podcast

This Week’s Article

54% of C-suite executives say AI is tearing their company apart.

That number comes from WRITER’s April 2026 Enterprise AI Adoption Survey, which surveyed 2,400 knowledge workers, including 1,200 C-suite executives and 1,200 employees. WRITER also reports that 56% of executives say AI has created power struggles and disruption inside their organizations, while 97% say AI has benefited them personally and only 29% say their organizations are seeing significant ROI from generative AI.

The finding is already showing up in broader coverage of workplace AI backlash. Business Insider used the same WRITER/Workplace Intelligence finding in a May 2026 piece on companies pushing AI adoption while citing AI for job cuts. Axios covered the prior-year version of the finding in March 2025, reporting that AI adoption was deepening divisions and creating new power struggles between executives and workers.

The easy read is obvious: crisis. Companies moved too fast. AI got ahead of the organization.

I don’t think that’s the most useful read.

The 54% number matters because it points to something deeper than adoption pain. AI is starting to expose the parts of the organization built around scarcity, delay, control, and permission.

When executives say AI is tearing their company apart, they’re describing a fight over control: decisions, information, approvals, budgets, staffing, and the authority to define what counts as true.

The same survey gets more specific: 56% of executives say AI has created power struggles and disruption inside their organizations. “Power struggles” is the important phrase. AI creates friction when it reaches the systems that govern how work moves through the company: forecast ownership, customer knowledge, approval chains, staffing models, performance standards, and management authority.

A forecast long owned by finance can now be challenged by model-assisted inputs from sales, operations, merchandising, or customer support. Customer insight concentrated inside research becomes queryable by product, marketing, and frontline teams. Managers who created leverage by routing information have to create leverage through judgment, prioritization, and accountability.

Every serious change in work changes who gets control. A decision becomes a workflow. A process becomes a query. A function that owned the answer becomes one input in a wider system. The people who built status, budget, and authority around the old version of the work feel the change first.

That’s where the friction starts.

The companies feeling the most internal pressure are the companies where AI has reached the operating layer: the layer where work is assigned, evaluated, approved, funded, and defended.

Now consider the other 46%, the companies reporting a quieter rollout.

Some are early. Some are disciplined. Some really are sequencing AI adoption thoughtfully. Many are doing something else entirely: mistaking visible AI activity for organizational change.

They have pilots. They have internal demos. They have productivity tools. They have executive enthusiasm. They have employees quietly using AI on work they’d never formally submit to a workflow review. They have a chatbot, a copilot, an innovation council, and a handful of real examples from employees who are getting useful work done faster than before.

The activity is real, and in pockets, so is the change. Someone in marketing is producing more variants. Someone in support is finding patterns faster. Someone in sales is summarizing accounts before a manager asks for a brief. Someone in operations is using AI to see exceptions that used to hide in spreadsheets.

Those gains matter, but they also create a new problem. Eventually, the company has to decide what those gains mean.

If one analyst can do the work of three, the staffing model has changed. If a frontline manager can query customer patterns directly, the ownership of insight has changed. If a model can turn a weekly report into a live operating signal, the meeting structure has changed. If approval work can be embedded into a workflow, the authority of the approver has changed.

Many companies avoid that conversion. They let AI change work informally while keeping the formal system intact. The old org chart remains. The old budget logic remains. The old approval chains remain. The old claims over data and expertise remain. The company gets more output from individuals while postponing the question that creates organizational value: who controls the work now?

That’s the quiet-company problem. Calm can look like maturity. Often, it’s just unabsorbed change. The organization has activity at the edge and avoidance at the center.

AI can create individual leverage long before the organization captures business value. That’s the ROI gap, and the survey captures it in two numbers: 97% of executives say AI is benefiting them personally. Only 29% say their organizations are seeing significant ROI from generative AI.

That is the enterprise AI problem: individual gains are showing up everywhere, while organizational returns remain stubbornly hard to capture.

Individual benefit is easy to understand. Faster drafts. Cleaner synthesis. More leverage. Better preparation. Less time spent staring at a blank page or rebuilding a deck from scratch. Executives feel those gains immediately because AI makes their own work easier.

Organizational ROI requires a second step.
A company has to convert individual leverage into operational change. That means deciding which workflows should be redesigned, which decisions should move closer to the work, which approvals should disappear, which teams should lose exclusive ownership of the answer, which budgets should be reassigned to workflows, and which roles become harder to justify.

That conversion is where most AI efforts stall.
An executive can draft a decision memo in ten minutes while the decision still takes three weeks to clear the same committee. An analyst can produce a better forecast while the business still runs on the old weekly reporting cadence. A support team can spot customer patterns faster while the people closest to the problem still need three layers of approval to act.

Local gains prove the tools are useful. Operating changes turn that usefulness into ROI. Individual leverage creates the possibility; the operating model captures the value.

The ROI gap closes when control actually moves.

A company can celebrate faster output as long as old ownership stays intact. The fight starts when leadership asks what the new output changes about authority.

Finance may welcome a better forecast until that forecast starts incorporating standing inputs from sales, support, operations, and product.

Research may welcome better customer summaries until product and frontline teams get direct access to customer patterns that used to flow through research.

A manager may welcome faster status updates until status routing stops being enough to justify the role.

AI makes work faster, and it makes some claims of ownership harder to defend.

When control moves, someone loses the right to be the default owner of the answer. Someone loses a review step. Someone loses budget attached to a function. Someone loses status built around information scarcity.

That is the part executives underestimate.

People fight because control protects them from future changes they cannot see or stop. A lost review step can become a smaller team. A shared forecast can become a weaker claim on budget. Direct access to customer patterns can become a reduced mandate for the group that used to own customer insight. A manager who no longer controls information flow may have to prove value in a completely different way.

Given the broader anxiety around AI and jobs, that fear is rational. People have heard enough about automation, efficiency, and headcount reduction to understand where this can lead. A workflow change can become an org chart change. A faster process can become a headcount question. A new source of truth can become a new power structure.

No one has to announce that in the meeting. Everyone can feel it.

This is why AI transformation gets political so quickly. Companies redesigning work are also changing the protections people have built around their roles, budgets, and authority.

The companies that capture ROI are the ones willing to name that fear, design through it, and move control deliberately.

Microsoft’s turnaround is useful here because the company did more than adopt a new technology strategy. It changed what the company organized around.

Before Satya Nadella, Windows still had enormous internal gravity. It shaped priorities, protected assumptions, and defined how much of the company understood power. Microsoft had cloud assets, developer relationships, enterprise distribution, and technical depth, but the old center of the company still pulled decisions back toward the Windows era.

Nadella moved that center.

Cloud became the strategic core. Windows became part of a broader platform strategy rather than the organizing principle. Products that might once have been treated as threats to Windows became part of the ecosystem. Teams that had been rewarded for defending territory had to learn how to create leverage across products, platforms, and customer relationships.

That is what transformation looks like from inside a company.

The market sees a strategy shift. Employees feel a control shift.

The lesson for AI is straightforward: the tool is rarely the transformation. The transformation starts when the company changes what it protects, what it funds, what it rewards, and what it lets go.

AI will create the same kind of test for many companies. Some will add AI to the existing operating model and call it progress. Others will let AI force a harder question: what should become the new center of the company?

The companies that answer that question honestly are the ones with a chance to turn adoption into advantage.

“Where can we use AI?” is no longer the useful question.

That question is too easy. It produces pilots, training sessions, software licenses, and safe productivity stories.

The harder question is more uncomfortable:

What part of our operating model exists because information used to be slow, scarce, expensive, or hard to synthesize?

That question gets uncomfortable quickly.

Which decisions still require meetings because no one trusts the data? Which approval steps reduce risk, and which ones protect status? Which roles exist mostly to collect, translate, route, or repackage information? Which teams own answers that other teams can now reach directly? Which budgets are attached to functions rather than workflows? Which managers create judgment, and which managers create delay?

The answer will point at something someone owns.

A process someone built. A team someone manages. A budget someone controls. A title someone depends on. A version of the company that still has defenders because those defenders benefit from it.

That is why executives avoid the question.

A serious AI strategy eventually becomes a control audit. It asks which decisions, permissions, budgets, and accountabilities should move because the old constraints no longer apply.

The companies being torn apart are already feeling the audit start.

The companies that feel fine should ask whether they have conducted one at all.

There is a comfortable story executives can tell themselves right now: AI adoption is underway, employees are experimenting, productivity is improving, the company is moving carefully, the strategy is forming, and the organization is calm.

That story may be true and still be incomplete.
A company can have pilots, copilots, demos, training, executive enthusiasm, and real productivity gains while leaving the old operating model almost entirely intact. The question is whether AI has changed anything the company actually controls: who makes decisions, who owns information, who approves action, who gets budget, and who carries accountability.

A better test starts smaller: pick one meaningful workflow.
Pick something important enough to matter to the business. Map where the information enters, where judgment gets applied, where the decision lives, who has to approve it, which budget supports it, and who is accountable for the outcome. Then ask what AI changes if information becomes faster, cheaper, easier to synthesize, and easier to act on.

Can the decision move closer to the work? Can an approval disappear? Can the budget follow the workflow instead of the function? Can the team closest to the customer act without waiting for three layers of permission? Can a manager stop routing information and start owning outcomes?

When the answer is yes, control has to move, and the people affected by that movement will understand the stakes immediately. A decision moving closer to the work means someone else has less authority over it. An approval disappearing means someone loses a checkpoint. A budget following the workflow means a function may lose funding. Accountability moving to a new owner means an old owner may lose protection.

That is the work executives should be doing now. The 54% being torn apart are not automatically winning. Some will mismanage the transition. Some will confuse chaos with progress. Some will cut too deep, move too randomly, or let shadow AI create risks they cannot govern.

But a lack of tension shouldn’t automatically reassure anyone. Calm may mean the company has handled the rollout well. It may also mean AI has been kept away from the places where authority, budget, and accountability actually live.

The point is to find those places deliberately before the market, the workforce, or a more aggressive competitor forces the issue.

AI adoption tells you the software is in the building. Control movement tells you whether the company is actually changing.

Start there.