AI is giving businesses a new reason to remove layers of management. If AI can produce the reports, track the numbers and schedule the work, fewer managers seem to be needed.
In PE-backed businesses, that logic is increasingly finding its way into the Value Creation Plan as a cost line.
The savings can be real. But the case for them usually counts only part of what managers do.
The tasks AI can take on are easy to list: reporting, scheduling, monitoring performance, preparing first drafts.
Managers also do work that rarely appears in a job description. They settle decisions before they need to go higher. They chase the hand-offs between teams. They develop the people below them. And they notice early when something is starting to go wrong.
AI does little of that yet. When the layer goes, that work either moves somewhere else or stops being done.
Often it moves up. Decisions a manager would have settled now reach the executive team or the CEO, and they wait longer. Disagreements between functions that used to be resolved informally are escalated.
Sometimes nobody picks it up. Hand-offs slip, problems are spotted later, and nobody is developing the next generation of leaders.
Neither shows up in the business case for the restructuring.
At the same time, the AI programme makes its own demands on the same leaders. Processes have to be redesigned rather than simply automated. Someone has to oversee the tools and the decisions they make. People have to learn new ways of working while the business keeps running.
Bernard Marr expects “plenty of costly failures” where businesses layer AI onto processes designed for people. Deloitte’s 2026 survey of 1,000 change leaders found that 61% feel AI transformation is moving faster than their employees can handle, and 84% say staff are expected to take on transformation work on top of their existing workload.
The point at which a management layer is removed is often the point at which the remaining leaders have the most to carry.
Map what the layer actually does before taking it out. For each part of that work, decide whether AI will take it on, who else will do it, or whether it can stop.
Time the change to when the AI is working and in use, rather than when it is bought or launched.
Then check afterwards whether decisions and coordination have held up: whether more decisions are reaching the top, how long they wait, and whether hand-offs between teams are slipping. These are early signs, visible long before the effect reaches the numbers.
Execution Radar treats an AI programme like any other initiative in the plan, reading what it demands of the organisation and whether the organisation can carry it alongside everything else.
We help investors and leadership teams identify execution problems, understand their causes and make changes that improve delivery.
