Rightworking, Not Rightsizing

Rightsizing was always a euphemism, and everyone knew it. It dressed a headcount cut in the language of correctness, as if there were an objectively right number of people and the organization had simply drifted above it. The word did real work. It made reduction sound like precision. But underneath the vocabulary was a single question, asked over and over for forty years. How few people can we get away with? Rightsizing optimized the denominator. It took the work as fixed, and the workforce as the variable to be minimized.

AI is now tempting every organization to do rightsizing again, only faster. The question on the table in most boardrooms — which jobs can AI take over — is the old question wearing a new tool. It assumes the work is a fixed pile of tasks, and that the exercise is to decide which of them move from a person to a machine. It is the most natural question in the world. It is also the wrong one.

It is wrong for a structural reason, not a sentimental one. AI does not do human work faster. It does different work to reach the same outcome. A machine asked to keep a condition true — such as new hires productive quickly, or people paid accurately, or roles filled with the right capability — does not perform the human steps more efficiently. It dissolves those steps and reaches the outcome another way. So the task list you were about to redistribute is not a fixed pile. It is an artifact of how that outcome used to be pursued, under constraints that no longer bind. Start from the tasks, and you have already smuggled the old model into your answer. You will automate your way to a cheaper version of the wrong thing.

So start somewhere else. Start from the outcome.

Rightworking begins with the condition you are responsible for keeping true, and asks a different question. What division of work between people and AI actually makes this outcome true, more often, and more timely? Design that division honestly, and what remains for the human is not the residue of automation. It is a redefined role, and much of it is new. Exercising the judgment the machine surfaces. Handling the exceptions it escalates. Governing the system. Carrying the trust and context a machine cannot. The human share is not what is left over. It is what is worth a human doing, once the machine is doing what it is good at.

This is where rightworking parts company with rightsizing, and the difference is not cosmetic. Rightworking does not always point down. Sometimes the right design frees human effort, and you move it to where it is needed more. Sometimes it keeps the same people and produces a dramatically better outcome — the same effort for a truer and faster result. And sometimes the right answer is to invest more human effort in an outcome, because it matters, and because people, now amplified, can move it as they never could before. Three directions, not one.

Headcount stops being the target and becomes an output. You never aim at a number.

That is also what keeps the word honest. Rightsizing corroded trust precisely because everyone could see the number was decided first, and the analysis reverse-engineered to reach it. Rightworking earns its name only because it is capable of coming back and saying: keep these people, or add, and get far more for it. A method that can only ever conclude fewer is rightsizing in a better suit, and a workforce will smell it in a week.

Behind all of this is the shift the AI era actually demands. The value on offer is not cheaper work. It is outcomes that are true more often and sooner, and the freedom to point scarce human judgment at the things only human judgment can hold. Not more output from fewer people. More capability, more reliability, more timeliness, from people doing better work. More with more.

The method is deliberately plain. Name the ten or twelve outcomes a function exists to keep true. Estimate, from the organization’s own leaders, how much human effort each takes today, and how well it is actually achieved. No employee time studies. No surveillance. For each outcome, design the sensible division of work between people and AI, shaped by the organization’s real constraints on what it can safely let a machine do. Estimate the redesigned state — the human effort that remains, including the new work people take on, and the outcome performance the design would deliver. Then compare, and let each outcome sort itself into effort freed, more with more, or invest to win. No benchmarks. The numbers come from designing the work, not from industry averages no one trusts. No proprietary data leaves the building. A strategic estimate, in ranges, validated by the people who run the function.

We call the engagement that does this a Rightworking Analysis. It produces, outcome by outcome, an honest picture of where human effort goes, what the right design would change, and where the real value is hiding. Usually not in cutting, but in redeploying judgment toward the outcomes that were never getting enough of it.

The deeper point is simple. You cannot cut your way toward a model whose whole premise is that the work itself is about to change. Outcome by outcome, you design the right partnership between people and machines, and you let the size of the workforce be the consequence of good design rather than its goal.

You do not rightsize your way into the AI era. You rightwork your way in.