The Frontline Paradox

There is a person in every retail business who knows more about the customer than anyone in the building where the decisions get made. She works Thursday nights and Saturdays. She knows which regulars come in for the conversation as much as the product, which promotion confuses people at the shelf, which step of the returns process makes customers sigh. She has never been asked about any of it. And when the new system arrives, she will be the first person measured on how well she uses it.

That is the frontline paradox. The people closest to the customer are the last to be asked and the first to be measured.

I do not think anyone designs it this way. I certainly did not, in the years I spent in head offices deciding things for stores I visited on Tuesdays. It happens because of distance, and because of the way information moves in a retail organisation. Insight flows up slowly, through layers, losing its edges at each one. Instruction flows down fast, on a Monday morning, with a deadline. By the time a store team's experience of the customer reaches the room where the technology is chosen, it has become a bullet point in a deck, if it has arrived at all. By the time the technology reaches the store, it has become a target.

What it costs

The cost shows up in a pattern I have now seen enough times to expect. A capability is bought. The pilot goes well, because pilots are run by the people who chose the capability and want it to succeed. The rollout happens. And then the people in stores, who were not asked whether the thing solved a problem they have, quietly work around it. Not out of resistance, or not mostly. They work around it because it was designed for a version of their day that does not exist, and they still have to get through the actual day. The scanner that takes four taps where the old way took one. The dashboard that tells the store manager what she already knew at seven this morning. The AI recommendation that is right about the customer eighty per cent of the time and embarrassing the other twenty, in front of the customer.

MIT's 2025 study of enterprise AI pilots found that ninety-five per cent delivered no measurable return, and the reason was rarely the technology. It was the organisation around it. Gartner's January 2026 analysis lands in the same place: at least half of generative AI projects abandoned after proof of concept, for organisational reasons rather than model ones. I would put it more specifically. It was the distance between the people who chose the technology and the people who were expected to carry it.

Why asking is harder than it sounds

Every leader I know would say they want frontline input. Most of them mean it. Asking is harder than it sounds for reasons that are quite human. Asking implies you might hear something you cannot act on, and then you have to go back and say so. Asking takes time you do not have in a project plan that was built around a vendor's timeline. Asking creates an expectation. And, if I am honest about my own years in head office, asking risks learning that the problem you have chosen to solve is not the problem the store has. That is an uncomfortable thing to learn after the contract is signed.

So the asking gets replaced by things that look like asking. A survey with questions written by the project team. A pilot in the stores most likely to say yes. A "champion" in each region, selected for enthusiasm. None of these are bad. None of them are the same as sitting in the back room at ten past six and listening.

Trust runs in both directions

When we built GRIT, our model for human readiness, we gave one of its four dimensions the name Frontline Trust, and we were careful about the wording. It is not whether the frontline trusts the technology. It is whether the people in stores believe that the data collected about their work is used to help them rather than to catch them, and whether the organisation trusts them enough to ask before it measures.

Both directions matter, and they are linked. A frontline that is measured before it is asked learns that the measurement is the point. It learns to perform for the dashboard. It does not learn to tell you that the process is broken, because nobody has ever shown that telling you leads anywhere. And so the organisation loses the one source of information that would have made the technology work.

The paradox resolves, in the businesses I have watched resolve it, with a change in sequence rather than a change in values. Ask first. Measure second. Show the frontline what you did with what they told you, especially when the answer was "we could not do that, and here is why". Let a store manager stop a trial that is not working without it counting against her. Put the people who serve the customer in the room where the problem gets chosen, not just the room where the rollout gets planned.

None of that is expensive. All of it is uncommon. If you would like to know where your own organisation sits, the GRIT assessment asks sixteen statements, written from the shop floor, and gives you a state for each of the four dimensions. We expect the Frontline Trust score to be the one that surprises leadership teams most often, and the surprise is the useful part.