“We have done plenty of this at Regal Rexnord and the results are impressive,” Dickson said. “Our internal GPT handles around 2,000 associate queries a month, the website chatbot helps more than a thousand customers a week find what they need, and the thousands of people using Copilot tell us they save two to three hours a month — with most of them putting that time straight back into higher-value work. Those outcomes represent genuine improvements, both for Regal Rexnord and our customers.”
The question is whether AI can be woven into the end-to-end processes that actually run a business — forecasting, demand planning, sales and operations planning, materials requirement planning, the customer service workflows that stretch across half a dozen systems and just as many teams. These are not chatbot use cases. They are the load-bearing operations a manufacturing company leans on every day, and running AI inside them is a different challenge entirely. It calls for multiple agents working together rather than in isolation. AI needs to be connected to core systems rather than bolted on beside them, and for those agents to be governed properly — monitored, observable, and with a human expert accountable and responsible for what they do.
“None of that is trivial,” he said. “But it is precisely where the transformation happens, because the moment you stop using AI to help people do their existing work faster and start using it to rethink what the work is, you have crossed from using a tool to developing a new capability.”
Why Most Organizations Stall
There is a pattern that repeats itself in industrial AI. A company runs a few successful pilots, generates real internal excitement, and then watches that momentum evaporate when it tries to scale. The technology works and the pilots prove out, yet the leap from a working pilot to AI genuinely embedded in core operations takes far longer than anyone promised at the start. It is worth asking why this happens, because the answer is rarely the technology itself, Dickson said.
The unglamorous foundations have not been built. Three tend to be missing. The first is data readiness. AI is only as good as the data that fuels it, and most industrial businesses still have data spread across systems that were never designed to talk to each other. Treating data as a product — with governance, ownership and clear definitions — sounds bureaucratic, but it is what makes enterprise-scale AI feasible.
The second is governance. With AI capable of touching important processes, the question of what an AI model is allowed to do, on whose authority, and with what oversight becomes a real one. Organizations that have not thought this through in advance tend to discover the gap at the worst possible moment, sometimes with disastrous consequences.
The third is change management. AI transformation is fundamentally about people, and about whether the people doing the work feel that the change is being done with them or to them.
The most effective way to deploy AI is as a tool to enhance experts’ skills, not as a replacement for them. Bringing everyone along is often the difference between a pilot that scales and one that stalls.
Transformation is Not a Project
The key mistake businesses need to avoid is thinking of AI transformation as a finite program with a start and an end. It is not. The technology will continue to evolve, the use cases will continue to expand, and the boundary between what humans and AI agents do best will continue to shift.
Treating it as an ongoing transformation looks different. It means building organizational capabilities — in data, in governance, in upskilling, in change management — that can absorb successive waves of AI technology as they arrive, rather than scrambling to react to each one. It means investing seriously in the people who will be using these tools, not as an afterthought but as a deliberate part of the operating model.
The Next Three Years
