I was at a manufacturing company last year. Beautiful new AI dashboard. Twelve screens. Real-time data flowing in from three systems. The kind of setup that looks great in the implementation deck.
I asked the floor supervisor if she used it.
She looked at me the way you look at someone who has just asked if you use the DVD player that came with the hotel TV.
The company had spent eight months and a meaningful amount of money deploying AI into their reporting layer. Clean outputs. Automated summaries. The executives loved the weekly briefing. The people doing the actual work had not changed anything they did on Monday morning.
This is not a technology failure. It is a placement failure.
The Layer That Gets Ignored
Every mid-market operation has a coordination layer. It is the unglamorous middle. Not strategy, not execution. The bit in between where information moves from one person to the next, where status gets chased, where documents get sorted, where the same question gets answered twelve times a week by someone who went to university to do something more interesting.
This layer runs on humans. And it does not have to.
I have spent the last two years walking operations in distribution, freight, construction, and professional services. The coordination tax is remarkably consistent. Somewhere between 30 and 40 percent of the working week, gone to work that predates the existence of AI by decades.
It does not show up in a budget line. It shows up in headcount. And in that low-grade exhaustion that mid-sized teams carry around like a laptop bag they have stopped noticing.
Insight
The most expensive coordination work in a mid-market operation is rarely the most visible. It lives in inboxes, WhatsApp threads, and the twelve-minute phone calls that happen before anyone opens a system.
Why AI Gets Deployed in the Wrong Place
The typical mid-market AI project starts in the conference room, not on the floor.
A founder reads about what a large company did. A consultant arrives with a framework. Someone picks a use case that sounds impressive rather than one that is expensive. And the AI lands somewhere visible rather than somewhere painful.
I understand why this happens. You have to show something. Invisible efficiency improvements are hard to photograph for the board deck.
But here is what I keep seeing. The AI goes into the reporting layer, the forecasting layer, the dashboard layer. The places where data already exists and outputs already get read. The coordination layer, where the actual daily cost lives, goes untouched.
Six months later, the AI is technically live. Nothing has changed on Monday morning.
What the Right Layer Looks Like
In a freight company we worked with, the bottleneck was intake. Documents arriving in seven different formats from sixty-three carriers. One person whose entire job was reading, extracting, and re-entering. Not because that person was not good. Because nobody had ever mapped that step and asked whether it needed a human.
We mapped it. It did not.
Four weeks to build the AI layer. The person moved to exception handling. Carrier onboarding dropped from eleven days to three.
In a professional services firm we worked with, the bottleneck was pre-meeting research. Consultants spending forty to sixty minutes before every client call pulling together account history, recent news, open actions. Not analysis. Retrieval.
We built a system that did it in ninety seconds.
Neither required a new platform. Neither required replacing a system. Both required someone to stand in the coordination layer long enough to see what was actually happening there.
Worth noting
The impulse to deploy AI across multiple workflows at once is the most reliable way to deploy it well nowhere. One workflow. One team. One measurable outcome. Then the next.
The Question Worth Starting With
Not "what AI can we add" but "where is coordination eating the most time this week."
That question has a specific answer inside every operation. It is usually not where the company thinks it is. And it is almost never on the dashboard.
If you are trying to find that answer, that is exactly what the Discovery Audit is designed to surface.