Last quarter, an AI drafted a 40-page change plan for an ERP rollout in 90 seconds. It was good. Better than some I have paid consultants five figures for. That was the moment I stopped asking whether AI would touch my discipline, and started asking what is left for me to do.
I lead AI and ERP change management at Novartis. I also build AInspire, a platform that reads engagement signals across transformation programs. So I sit on both sides of this: I use these tools, and I ship them. Here is what I actually see happening.
What AI already does better than us
Three things, concretely.
Drafting. Stakeholder maps, communication calendars, training curricula, FAQ banks. The blank-page tax that used to eat the first two weeks of a program is gone. I now start from a solid draft on day one and spend my time editing judgment into it.
Sensing sentiment. We used to run a pulse survey, wait ten days, and read averages. Now models parse Teams channels, support tickets and meeting transcripts in near real time. On one rollout, sentiment analysis flagged a demoralized regional team eleven days before the survey would have. We intervened while it still mattered.
Predicting resistance. Given adoption curves from past programs, models forecast which groups will stall and roughly when. Not magic — pattern matching on history. But it turns resistance from a surprise into a schedule.
The uncomfortable truth: 60 to 70 percent of what change managers were paid to produce is now a prompt. Pretending otherwise is how a profession gets disrupted from underneath.
What does not transfer to a model
The draft is not the change. It never was. A model can tell me a factory floor will resist a new MES workflow. It cannot walk onto that floor, notice the shift supervisor everyone actually listens to, and win her over before lunch.
Here is what stays human — and gets more valuable, not less:
- Trust. People adopt change from people they believe. An AI-written apology after a botched go-live is worse than silence. Presence cannot be generated.
- Political judgment. Which battle to fight, which executive to let save face, when to slow down a "successful" rollout because the organization is quietly exhausted. Models optimize; they do not read a room.
- Accountability. When a transformation fails, no one accepts "the model recommended it." A human owns the decision. That ownership is the job.
- Meaning. AI can predict resistance. Only a person can stand in front of 200 anxious employees and make the change feel like something worth doing.
The role is inverting
The old change manager was a producer: decks, plans, comms, artifacts. The new one is an editor and an operator. You spend less time making the plan and far more time in the field — testing it against real humans, catching what the model got confidently wrong, and acting on signals faster than any committee could.
The math is stark. If AI removes 60 percent of my production work, I have two options. Deliver the same programs with a smaller team — which is happening — or take the reclaimed time and go deeper on the human layer that no model touches. I choose the second, and it is where I now create the most value.
What I would tell a transformation leader hiring today
Stop screening for people who can build a change plan. AI builds change plans. Screen for the person who can tell you when the plan is wrong, sit with a furious plant manager, and still land the outcome. Those skills were always the hard part. Now they are the only part worth hiring for.
What I would tell a change manager reading this
Do not compete with the model on speed or volume. You will lose. Compete on the things it cannot fake: presence, courage, judgment, care. Let AI take the 60 percent that was never really the work. Then become undeniably good at the 40 percent that always was.
AI did not come for the change manager's job. It came for the parts of it that were never really change management — and quietly handed us back the parts that always were. The question is no longer whether to use these tools. It is whether we are brave enough to do the human work they can never do for us.
Cédric Bignet is an AI & ERP Change Management expert at Novartis and founder of AInspire. He writes about change management, AI adoption and enterprise transformation.