By Steve Urban
Most of the AI conversation in business still happens at the top. Board decks, strategy offsites, a new title on the org chart. Meanwhile, the people closest to the work are quietly getting on with it.
A new survey put some numbers to what a lot of operators already suspect. The IFS Industrial Index 2026, covered by DC Velocity in early November, surveyed 2,275 executives and frontline workers across 14 countries. Its headline finding was that engineers, plant managers, service leads, and logistics managers are the ones driving real AI use, using their own operational data and what they know about the work. More than 90 percent of organizations reported significant gains in reliability and performance.
The leadership side looked different. Executives said their top AI effort was building their own AI literacy. Leadership confidence showed up as one of the biggest barriers to scaling what was working. And frontline progress was being slowed by governance rules and data access that had not kept pace.
IFS sells industrial software, so read the gain numbers with that in mind. But the pattern matches what I have seen for most of my career, long before anyone was talking about AI.
I have spent more than 30 years running operations, from retail districts to consumer products companies to the companies I run and advise today, including MPS Security, Riderflex, Robo Reliance, and Kura Home Services. In almost every one, the most useful improvements came from someone who did the work every day and got tired of a stupid step.
A supervisor who knew which report nobody read. A rep who had built a better tracking sheet than the official one. A tech who knew exactly which part failed first. Leadership’s job was usually not to invent the fix. It was to notice it, make it safe to try, and then spread it.
AI does not change that. It speeds it up. The person who knows the workflow is the person who can tell in five minutes whether an AI tool helps or just creates more cleanup.
I want to be fair to the other side, because it has a real point.
Leaders are right to worry about people pasting customer data into whatever tool they found last weekend. They are right that a dozen disconnected experiments can create a mess that is expensive to clean up later. And they are right that someone has to own the risk when an AI-assisted decision goes wrong. Even IFS warns that scaling AI agents without oversight creates risks like unvalidated outputs, data exposure, and decisions with no one accountable.
So the answer is not to let everyone do whatever they want. Guardrails matter.
The problem is when caution turns into a permission process slower than the problem it is supposed to manage. If it takes a quarter to approve a tool that saves a dispatcher an hour a day, the business has quietly decided that the hour does not matter. And the best people will either find a workaround or stop trying.
Here is what I would focus on if I were walking into a company today:
Go find out what is already happening. Ask your frontline leads what AI they are using, officially or not. You will learn more in a week of honest conversations than in a month of vendor demos.
Make the safe path the easy path. Approve a short list of tools, write down in plain language what data can and cannot go into them, and get out of the way.
Fix data access. A lot of frontline AI stalls because the person with the idea cannot get to the data. That is a leadership problem, not a technology problem.
Give the owner a name. Every AI use that touches a customer, a dollar, or a safety decision needs one person who owns the result. Not a committee.
Spread what works. When something saves real time on one team, put the person who built it in front of the others. Peer proof beats a mandate every time.
Learn by doing, not by course. AI literacy for executives is fine, but the fastest way to learn is to sit with the people using it and watch.
There is a lot of talk about AI replacing management layers. I do not think that is the right worry. The real risk is a leadership team that is busy learning about AI while the operation is already ahead of them, and then slows it down because it does not understand what it is seeing.
The leaders who get the most out of this will be the ones who treat frontline AI the way good operators have always treated frontline ideas. Listen first. Set clear rules. Remove the roadblocks. Then hold people accountable for results, not for asking permission.
Your people are probably further along than you think. The question is whether you are helping them or standing in the doorway.
Connect with Steve: linkedin.com/in/stevepurban
Photo: Shavr IK on Unsplash