It’s just another day on your factory floor. One of your technicians is crouching next to a vital machine, juggling a handful of tools and trying to wrangle a glitchy app at the same time. Meanwhile, your managers and executives are poring over glossy dashboards in nearby offices and a distant boardroom. But the critical machine remains down.
We hear constantly about AI-enabled manufacturing and autonomous manufacturing that promise to make production a breeze. So why does the reality on the floor often feel different?
Here’s the irony: AI promises us a new industrial revolution. Yet adoption often falls short of expectations.
THE FLAWS IN TOP-DOWN AI DEVELOPMENT
That’s because well-meaning managers and executives start with their own ends in mind, creating a top-down approach to problem solving. That’s rarely the most effective way to do it, though. It creates a familiar trap: costly executive dashboards that look impressive but are rarely used and deliver little real value True value emerges by empowering frontline workers first, allowing their insights to bubble up organically. That leads to software built for those technicians, which drives better data and operational outcomes. Think of it as flipping the script, starting with the wrench (technician tools) rather than the dashboard (executive views).
That’s important because the typical knee-jerk development of AI tools designed for executives and analysts are driven by boardroom KPIs like uptime or throughput. But they rarely incorporate input from floor workers about the daily problems that impede their work, so they fail to drive lasting change. Decisions made in office meetings lead to software purchases for top-of-the-house visibility. But if your techs don’t use it, then the data collected is often smoke and mirrors.
KPIs developed without frontline feedback miss real-time issues and ignore daily challenges like machine breakdowns or changeover times, losing valuable insights. You wind up with software that functions like static reports which is great for meetings, but disconnected from what’s actually happening on the factory floor. Real AI value requires clean, actionable data straight from the source. The top-down approach stifles adoption, wastes resources, and fails to address operational realities, leading to stalled or pointless digital “transformations.”
WHY START WITH FRONTLINE WORKERS?
Building your AI solutions for your technicians first ensures usability and buy-in. Empowered techs can input high-quality data naturally, leading to better reports and AI insights bubbling up to all levels of the organization. If the software fits your workflows, technician adoption will soar. For example, data can be entered by AI-enabled voice notes in “quick bites” right before breaks, versus on lengthy forms. And more thorough adoption on your plant floor enables actionable metrics on big hitters like breakdown frequency and repair times.
Then combine operator input with sensor data for a full picture of your operations. Partner with tools for machine connectivity and prioritize human-facing software to avoid data silos.
AI tools that solve pressing, day-to-day problems in real time, like predictive alerts for techs, produce cleaner patterns for use throughout the organization.
STRATEGIES FOR BOTTOM-UP AI ADOPTION
Now here’s a counterintuitive reality many teams discover: Top-down AI is easier to implement. Or at least, it seems easier to implement in the early days. So it can lend itself to a short-term victory declaration that over the long haul, usually leads to disaster.
Do things the right way by tackling the tougher path of bottom-up implementation with an incremental approach.
Start small to avoid data indigestion: You don’t have to instrument every machine at once, for example. Instead, instrument your most critical machines first. Prioritize the bottleneck on each of your systems as Lean principles advise. That should be your floor technicians’ focus anyway, right?
Avoid waterfall integrations with their rigid critical paths and timelines: You need flexibility to ensure early portions of your implementation are working before you move to later phases and divert the focus from those critical first stages. Instead, build your end-to-end flows gradually.
Help your tech help you by evolving with it: Use cloud-based software for rapid updates, integrating with your existing systems like your ERP and MES for a complete stack, including human input, machine data, and production and quality metrics.
Involve your users at every implementation stage: Develop your new systems and practices with direct technician feedback to ensure the tools make their jobs easier, and they’re not just demos that “look cool” for 20 minutes.
Know up front that there will be resistance, since you’re likely asking highly experienced floor technicians (your true experts) to do at least part of their jobs differently. Overcome that resistance by fostering bottom-up adoption over mandates from the get-go. Prove to your experts with your early results that good data entry naturally yields not only meaningful executive reports, but a better-running operation and a happier place to work.
FINAL THOUGHTS
Successful factory AI adoption means flipping the script and doing it bottom-up. Starting with the wrench, not the boardroom, unlocks real value by empowering technicians, surfacing frontline insights, and avoiding costly tools that look impressive but go largely unused.
Your call to action is this: Urge your manufacturing technicians and leaders to cooperate with your developers to rethink priorities, focusing on field usability for sustainable gains. Not only will you have quick wins with current technology offerings, but the long-term impact compounds. As sensors and AI advance, bottom-up strategies will drive a competitive edge in a fast-evolving industry. That willingness to adopt new ways of doing things will continue paying dividends as your frontline experts seek new technical solutions.
In the end, you’ll know it’s a success if everyone is satisfied, top to bottom. If the technician loves it and puts great data in there, the executive will love it at the end too.
Hugo Dozois-Caouette is CTO and cofounder of MaintainX.