Under project DESET, we have just finished a study that goes after something other than opinion. Instead of asking people what they think, we collected small changes that had already taken place in their companies. The method was SenseMaker, a narrative approach we have used often enough by now to trust. Twenty-five stories came in, all from manufacturers, most of them exporters. People told them in their own words and then interpreted their own story. Reading through them, one thing stood out. The small changes in these companies rarely start with the owner or the managing director. They start with someone on the shop floor, or in middle management.

That is worth pausing on, because the people filing these stories are mostly owners and managers. Ask them who deserves the credit and they point away from themselves. The change is usually carried by technical staff: a shift supervisor, a production technologist, an engineer, with the line technician close behind. Owners rarely cast themselves in the lead role. They can see where the change is actually coming from.

The stories show why. At one company a mechanical engineer designed a long rotating conveyor that turns the bottles and cools them as they go, and it works remarkably well. No outside help of any kind. At another, a machine went down and an operator bypassed the sensors, running it on experience until the replacements arrived. There are more like that. Nobody told any of them to do it.

Improvisation of that kind is also the most fragile thing in the sample. Where the change is carried by the people operating the machines, it has the shortest average life. Where technical staff or management carry it, it lasts noticeably longer. A good idea that lives in one head walks out the door with the person. Or it fades once the first burst of enthusiasm is over. The rotating conveyor runs for as long as the man who thought of it is still there. Bypassing the sensors bought the plant a few days of production, but nobody can inherit that. It is a trick, and tricks get forgotten.

What survives is what has been given some formal shape. The data say so: a change that has been formalised, turned into a rule or a procedure, lasts about two points longer on a ten-point scale than one left informal. Two points is a lot here. It is roughly the gap between a change that holds and a change that disappears. Improvisation is a good source of ideas. It is a poor way of keeping them.

None of this means the answer is to write everything down. Shop-floor people generally dislike paperwork, and several stories say so plainly. A form looks to them like extra work nobody asked for, slowing down the job they are paid to do. Formalisation also tends to be read as an order from above, which brings on exactly that resistance, and resistance clips the reach of a change even when the change is a good one. The task is not to pin ingenuity down in a procedure at any cost. It is to do it so that the person who came up with the solution gets something out of it.

So what can be done in practice? Alongside everything else they do, projects like DESET can take on one modest job: helping a company write down what it already knows how to do. Not a new machine, not a grand plan. A ready-made form, an hour of training, a maintenance card, a measuring routine, all pitched low enough that they do not feel like a burden. If AI can take some of the weight off the data entry and cut the resistance that comes with it, better still. That is AI doing something genuinely useful. For the effort involved, this is probably where the return is highest, because nothing new has to be invented. Something that has already proved its worth is simply kept from being forgotten. And since stories with an outsider involved tend to do better, while half of them had no outsider at all, there is a second job of equal weight: lowering the barrier that keeps a company from asking for outside help in the first place.

The encouraging part is that nearly every company in the sample is planning a next step. Ideas are not in short supply. The shop floor turns them out on its own. What is missing is a form that lets them outlive the shift they were born in. If the study yields one usable thought, it is probably this: the most valuable thing we can give a company is sometimes not a new idea, but making sure its own good idea is not left to evaporate.

* This blog was translated from Serbian using AI. Original version can be found at https://edabl.org/home/