By Nick Haase, Co-Founder, MaintainX
Enterprise IT has funded digital transformation for years, and most of it still underperforms, because the solutions get designed around systems instead of around the way people actually work. AI amplified that problem instead of solving it. Manufacturing made the gap impossible to miss: IT and plant-floor teams operating on different priorities, different assumptions, and no shared view of the problem either one was trying to solve. The pattern in the deployments that work is consistent. Balance security against ease of access, feature depth against a clean experience, and never break the workflow already in motion. The lesson for enterprise IT is straightforward. Treat every environment like a plant floor and build around the people closest to the work.
Somewhere in a plant right now, a technician is writing on a clipboard three feet from a tablet IT rolled out last spring.
That clipboard is a vote. It is the most honest feedback an enterprise technology strategy will ever get, and almost nobody who approved the deployment will ever see it.
Ask a CIO how the last major rollout went and you tend to get the same answer in different words. It went fine on paper and quietly nowhere on the floor. That is not bad luck. It is the default outcome. McKinsey puts the success rate for digital transformation initiatives under 30 percent. AI is worse. MIT researchers found that 95 percent of generative AI pilots produce no measurable business impact. S&P Global reported that 42 percent of companies abandoned most of their AI initiatives in 2025, more than double the year before.
None of that is a budget problem. Global digital transformation spending is on pace to top $3 trillion in 2026. The money is there. What is missing is easier to name and harder to fix. Most of these projects are designed around the system instead of around the person who has to use it.
The carpeted side and the concrete side
Every plant has two sides. There is the carpeted side, where budgets get approved, architecture gets drawn, and vendors get scored. And there is the concrete side, where the work actually happens. The distance between them is where most technology investment goes to die.
Picture a mid-size plant during a routine equipment failure. A technician is diagnosing the problem and trying to get the line moving before the next shift walks in. Somewhere else in the building, often in a different time zone, an IT team has just finished deploying a platform built to prevent exactly this. The platform is secure. It cleared every compliance review. It looked sharp in the demo.
Then it meets the floor.
The technician cannot operate it with gloves on. The login was designed for someone sitting at a desk, not standing next to a running machine. Logging one repair takes eight taps across three screens. Adoption stalls. The clipboard comes back. Six months later, IT is asking why the data they need for reporting, let alone for training a model, was never captured in the first place.
The bill for that gap is not abstract. Unplanned downtime costs U.S. manufacturers an estimated $50 billion a year, according to Aberdeen. Siemens’ most recent analysis found that Fortune Global 500 companies now lose roughly 11 percent of annual revenue to unplanned downtime, nearly $1.5 trillion combined, up from 8 percent a few years ago. A meaningful share of that increase traces back to tools nobody on the floor wanted to use.
AI does not fix a broken foundation. It exposes it.
The rush to deploy AI made this sharper, not smaller. Predictive models are only as good as the data feeding them. If the system that was supposed to capture that data got rejected by the people doing the work, the model has nothing to learn from. You end up automating a gap instead of closing it.
Manufacturing does not tolerate confident guesses. People get hurt. Equipment gets destroyed. Downtime is measured in real dollars, not hypothetical ones. AI should make an experienced technician sharper and a new hire faster to ramp. It should not replace the judgment either one brings to a problem that does not fit a flowchart.
The strongest argument against all of this
Security and governance teams have a fair response here, and it deserves a straight answer. They are the ones who inherit the breach, the failed audit, the ransomware invoice. Loosening controls so a technician can log in faster is not a trade they are paid to make. They are right about the risk.
But a login airtight enough to satisfy an audit and clunky enough that a technician hands their badge and password to a coworker has not reduced risk. It has moved the risk somewhere nobody is measuring. Unusable security is not security. It is a compliance artifact with a workaround attached.
What the successes have in common
The organizations that got this right did not start by picking a platform and pushing it down. They started with the person closest to the work and built outward. In practice, that means holding three things in balance instead of optimizing for one.
Security and access have to coexist. Design the authentication for someone wearing gloves in a loud room, then harden it, rather than hardening first and apologizing later.
Depth and simplicity have to coexist too. Launching with everything the platform can do is tempting. Every extra field and every extra screen is a tax on someone trying to get equipment running again. The rollouts that stick launch narrow and add capability only once the core workflow is second nature.
New technology has to fit the workflow already in motion, not the reverse. Ask technicians to change how they move through a shift to accommodate the software, and most of them will quietly route around it.
Every environment is somebody’s plant floor
This is not a manufacturing lesson, and it is not really about maintenance. It applies to any enterprise team deploying anything new, AI included.
Find whoever is closest to the actual work. Design around them before you design around the dashboard. Launch narrow. Measure whether people open it without being told to. Add the rest later.
That does not mean IT hands over governance or lowers the bar on security. It means governance gets built with the frontline instead of dropped on it from three floors up, in a conference room nobody on the floor has ever seen.
The next wave of enterprise technology will not be judged by what it can technically do. It will be judged by whether the person closest to the work picks it up without being told to.
Design for that person first, and the adoption numbers, the data quality, and eventually the AI results tend to take care of themselves. Design for the dashboard, and you will get a very secure system nobody uses.
