The headcount question

A leader asks me, in roughly every fifth conversation, the same question. Phrased politely, it goes: given what AI will do to our headcount, how should I plan?

I have learned to ask back: what do you mean by "do to our headcount?"

This is not a rhetorical move. The question hides three different scenarios that require different answers, and the leader is usually conflating them, which is why the question feels stuck.

The first scenario is cost reduction. The company believes AI will let it do the same volume of work with fewer people, and is planning to translate that into headcount reductions and margin improvement. This is a defensible move in some businesses. It is rarely the right move strategically, because it assumes the activity is the right one to be doing, and locks in the current shape of the company at a slightly lower cost. It is, in essence, the workflow fallacy applied to org design.

The second scenario is capacity reallocation. The company believes AI will make some activities cheaper, and intends to redirect the freed capacity to activities the company is currently underinvested in. This is usually the right move, but it requires a clear view of which activities are underinvested, and that view is harder to articulate than most leaders think. Without it, "reallocation" becomes "people sitting around looking productive while waiting for the next assignment," which is the same outcome as cost reduction, just with worse morale and a delayed P&L hit.

The third scenario is shape change. The company believes AI will change which activities exist at all, and is planning to evolve the org around the new activity mix. This is the strategic answer. It is also the answer most companies cannot execute on in 2026, because they do not yet know which activities will exist a year from now. Pretending to know is worse than admitting they do not.

When a leader asks me about headcount, the first move is to figure out which scenario they are actually in. Most of them think they are in scenario two. Most of them are actually in scenario one, with the language of scenario two as a fig leaf. A few are trying to be in scenario three but have not done the diagnostic work to make scenario three real.

The reason this matters is that the wrong scenario produces predictable failures.

A scenario-one company that pretends to be in scenario two will free capacity, fail to redeploy it meaningfully, and lay people off six months later, having destroyed trust on the way. The org learns that "reallocation" means "we have not figured out the layoffs yet." Future change initiatives meet stiff resistance.

A scenario-two company that has not done the diagnostic work will free capacity into "innovation projects" that produce nothing, then lay people off, then watch the leaders of the redeployed teams quit, citing lack of clarity. The org learns that the leadership did not actually have a plan, and the people who could have led the next phase choose to do it elsewhere.

A scenario-three company that succeeds, on the other hand, looks confused for about a year and then looks unrecognizable. The first year is hard. The team is being asked to do unfamiliar work, against goals that are still being formed, while AI tools change month to month. Many people leave because the work is no longer the work they signed up for. New people arrive whose skills fit the new shape better. The org chart looks different. The roles look different. The output looks different.

I want to be honest about this last scenario. It is uncomfortable. Many of the people who were excellent at the old shape will not be excellent at the new one, through no fault of their own. The leader has a real obligation to be straight with people about what is happening, to invest in helping the ones who can transition, and to be generous and clear with the ones who cannot. I have watched leaders execute scenario three with grace and watched leaders execute it with cruelty. The first kind compounds. The second kind also compounds, but the wrong way.

The general advice I have arrived at, on the headcount question, is short.

Do not budget for headcount reduction in year one. The numbers people produce in year one are not stable. The activities that look automatable in January look harder in June, and easier again in December. Locking in a headcount target around an unstable estimate produces bad layoffs, bad timing, and a damaged org.

Do budget for capacity reallocation, but only in scenario two. And only if you have done the work to know where the capacity is going. "Innovation" is not a destination. "Tightening the eligibility decision" is. The first is a vacation. The second is a project.

Do communicate openly about scenario three when you are in it. People can handle "the work is going to change shape and we do not yet know what every role will look like" if it is said in those words and accompanied by visible investment in their development. People cannot handle "everything is fine, no plans for changes" followed six months later by abrupt restructuring.

Do not, under any circumstances, write the AI strategy memo that promises both efficiency gains and no impact on the team. That memo is a lie that people detect immediately. It corrodes trust faster than honest bad news.

A leader I respect framed this to me recently in a way I will steal. He said: "AI is not changing the headcount question. It is changing which questions about people are honest." I think that is right. The honest questions are no longer "how many people do we need." They are "what activities are we becoming a company that does, and which people are well suited to that?" These questions are harder. They are also the only ones that produce useful answers.

I have come to think the leaders who navigate this transition well are the ones who can sit, in the open, with not yet knowing the answer to the second question, while still acting on the first claim that the activities are changing. The leaders who insist on premature certainty about either piece are the ones who produce the worst outcomes for everyone, including themselves.

There is no shortcut to this. There is no readiness assessment, no benchmark, no consultancy that can tell you what your company is becoming. You have to figure it out, slowly, by working on the load-bearing decisions and watching what changes. The headcount question answers itself, in retrospect, once that work has been done.