Change Management Was Never Meant to Be Run from Spreadsheets

Most change programs are still being delivered with tools and habits that feel like they belong in the 90s. Excel impact assessments, stakeholder registers, PowerPoint status packs, readiness checklists, comms trackers, training matrices, risk registers, steering committee updates. The list goes on.

None of it is inherently wrong. Good change work needs structure and discipline. But after years of delivering change in energy, banking, government, retail, and health, I keep seeing the same thing. The profession has become extremely good at documenting change, and less good at successfully delivering it.

We classify stakeholders, color-code risks, summarize engagement, update packs, and report progress. Then we wonder why people still feel like change is being done to them. The problem is not that practitioners do not care. Most care deeply. The problem is that the operating model of change has not kept pace with the world it is operating in.

A half-million-dollar chatbot, two years later.

About two and a half years ago, I worked at one of the world’s largest professional services firms. We were selling a custom chatbot to a client as an add-on to a change program. The price tag for the chatbot alone was over half a million Australian dollars. At the time, that was a credible number for what it took to build.

Earlier this year, on my current program, I built a more useful version in a day, using Microsoft 365 Copilot. That is not a productivity story. It is a story about how fast the economics of change work are shifting with the progression of AI. The same shift is showing up everywhere I look. On a recent program, I cut the time to complete a full change impact assessment by roughly 40%. The method was to feed current-state and future-state design documents into AI, and apply the typical practitioner judgment to the output. That 40% probably understates the opportunity. The current-state documentation on that program was incomplete. With better source material, I am convinced that the saving would have been much higher.

From artifacts to signals

The interesting part of the chatbot story is not the cost. It is what the chatbot turned out to be useful for. My original thinking was the obvious one: give impacted users an easy way to ask questions. It does that. But the quieter value is what comes back the other way. Every question someone asks is a signal. It reveals confusion, training gaps, missed leader messages, or decisions people have not accepted. On my programs, I refer to this kind of chatbot as an AI Change Companion. Used well, it becomes a listening channel. That is the shift the profession needs to make. Change management has spent decades producing artifacts. The future is interpreting signals.

A readiness dashboard does not create readiness. A stakeholder map does not build trust. A status pack does not create alignment. What creates those things is what practitioners do with the signals sitting underneath them.

AI-assisted is not AI-led

There is a distinction worth drawing. AI-assisted change uses AI to make existing work faster. Better drafts, quicker summaries, cleaner reports. That is real value. Most change teams would benefit from it tomorrow. AI-led change asks a bigger question though. The operating model of change was built for human-paced analysis. What should it look like when AI handles synthesis and signal detection at machine speed?

This is the thinking behind GEROA™, a framework I developed and now use on programs. The letters stand for Govern AI, Embed AI, Reallocate human effort, Optimize with AI, and Accountable human oversight. It is not a lifecycle. It is a set of conditions for AI to strengthen change work rather than undermine it. Govern sets the boundaries. Embed builds AI into how change is delivered, not bolted on at the end. Reallocate is the human upside. The time AI gives back goes into stakeholder work, not new artifacts. Optimize uses live signals to adjust the change while there is still time. Accountability keeps human judgment on the decisions that matter.

What practitioners do with the time

The work AI should take away is not the work of judgment, trust, influence, empathy, or engagement. It should take away the manual burden that prevents practitioners from doing those things properly.

If AI cuts impact-assessment effort by 40% at a minimum, the question we need to be asking is not how to celebrate the saving. It is what do our teams do with the time.

I believe that as AI continues to advance, the future of change is not a faster deliverable. It is, ironically, a more human one. More engagement, more trust, earlier signals before they become resistance, and higher-quality change as a result. Because successful change has never lived in spreadsheets. It has always lived in the human work.