Why Most Continuous Improvement Programs Stall (and How to Fix the Loop)

Key Takeaways

• The goal of continuous improvement has not changed in 140 years. Only the speed has.
• Around 68% of shop floor problems never get reported, usually because reporting is painful and nothing comes of it.
• Most programs break in the same spot: the loop stays open, so the gains never hold.
• Digitizing paper is not continuous improvement. Closing the loop is.
• AI speeds up the rituals you already trust. It does not replace the data, the processes, or the culture underneath them.

Most manufacturers do not have a continuous improvement problem. They have a follow-through problem.

In a recent Poka webinar, James Gardner, who leads the presales team at Poka, and Phil Young, a digital transformation project leader with thirty years in the trenches, dug into why so many improvement programs look busy on paper but never move the numbers. The tools keep changing. The goal never does. And the place where most programs break is the same place it has broken for decades. The loop stays open.

What follows is the short version, plus the parts worth taking back to your floor.

The Goal Has Not Changed in 140 Years

Phil opened with a history lesson, and it earns its place. Continuous improvement was not invented by one company or one country. It was built by a handful of people handing off the baton across more than a century. Taylor gave us timing work and standardized tasks. Ford designed waste out of the assembly line. Deming carried statistical thinking to postwar Japan. Toyoda and Ohno built the Toyota Production System with just in time, Kanban, and the seven wastes. Shingo closed things off with SMED and mistake proofing.

Nobody even called it lean or continuous improvement until around 1990. The name showed up last. The thinking came first.

Phil’s real point was that the target has never moved: more output with the same or fewer resources. Run the Toyota Production System, run lean, run total productive maintenance, run an integrated work system, and you are chasing the same equation. The methodologies are one language spoken in slightly different dialects.

The thing that has changed is speed. The gap between each wave keeps shrinking. Decades passed between the early breakthroughs. Now Industry 4.0, integrated work systems, and industrial AI have all landed inside a single thirty five year window, and the wheel keeps spinning faster. Whatever tool you bring in should still sit on top of the rituals you already trust, root cause analysis and SMED and standard work included. It should make them faster, not replace them.

The Three Ways Continuous Improvement Breaks Down

This is the part Phil cares about most, and it shows, because he has lived all three.

The digitized paper toolbox

Plenty of teams tell themselves that moving a form off paper and into SharePoint or PowerPoint fixes the problem. It does not. You have changed where the checklist lives and kept the exact same broken habit. The paper stops being paper. Nothing else about it changes.

No closed loop

Teams are great at closing actions and celebrating wins. They are much worse at putting corrective and preventive measures in place and then holding the gain. A project wraps, everybody claps, and three months later the floor has quietly slid back to the old way of doing things. With no standard to sustain the change, and no history behind it, whatever you gained leaks away.

Guessing the problem

When the data is not there, people go on hunches instead. So you fix the same issue again and again, because you keep killing the symptom while the root cause sits untouched in the background.

When the audience got polled on which of these hurt most at their sites, one answer ran away with it. No closed loop. James said that is exactly what he sees walking into the biggest manufacturers in the world. Everyone is still trying to get there. Digitizing paper without a way to sustain the gain just hands you a faster way to lose ground.

What Guessing Actually Costs You

Phil put some benchmarks from Deloitte and the World Economic Forum on the screen. Take them as directional, not gospel, because both presenters were the first to say the real numbers are usually worse.

Illustrative industry benchmarks cited from Deloitte and the World Economic Forum. Both presenters noted real world figures tend to run longer.

James pushed on that last number, and not in a reassuring direction. Fourteen days, he said, is generous. Pulling the Pareto chart, wrangling a cross functional team together, waiting on engineering and quality to actually show up, that alone can burn the whole two weeks before anyone lays a hand on the real problem. An undocumented fix is simply how the same problem finds its way back to your floor.

The 68% Problem Nobody Talks About

Then came the number that should stop you cold. Roughly 68% of problems, defects, and improvement opportunities on the shop floor never get reported.

Not because your people do not care. Because reporting is a pain. An operator raises a hand once. Nothing happens. So they stop. The continuous improvement culture you keep preaching quietly dies, because raising something led nowhere.

The systems often make it worse. A machine throws an emergency stop, and the Pareto chart or the manufacturing execution system logs the E-stop with no context at all. Why did it stop? Did a door open? What was the actual cause? None of it gets captured. So the operators stop looking, the data quietly rots, and the whole team drifts into firefighting, chasing the loudest problem of the shift instead of the one actually choking throughput at the bottleneck. Make reporting hard and let it lead nowhere, and people stop reporting. It is that simple.

Where Continuous Improvement Succeeds

The pattern behind every strong program is not a mystery. Capture the issue. Analyze the data so you are chasing the right one. Find the root cause. Put the improvement in. Close the loop and train on the new standard. Then celebrate the win, and the people behind it.

That last part is the one that gets skipped. Teams rush to celebrate the result and forget the person who made it happen. James kept coming back to the little wins. A near miss that got reported. A safety issue flagged. A 5S board someone bothered to fix. Give the operator who caught it a high five, because people drive change, and celebrating the small stuff is what gets them wanting to do more of it.

He and Phil both had stories, but Phil’s stuck. An operator told him a 5S cleaning board was in completely the wrong place. Instead of overruling him, Phil asked why. Turned out the crew kept the brushes and shovels wherever they actually used them, not back on a board bolted to a wall in the wrong corner. Small thing. Real reason. And the kind of buy in you cannot manufacture any other way.

Where AI Fits, and Where It Does Not

Both presenters were careful here, which matters with an audience that has been burned before.

AI does not replace the rituals. It speeds them up. The find and fix loop stays exactly the same, capture through analyze, root cause, fix, prevent, sustain. What changes is how fast you get through it and how much context you carry into each step.

A few things stood out about how that plays out inside Poka:

• It runs on your own production data, not the open internet, so every recommendation traces back to your equipment, your standards, your history.
• It is governed and auditable. You can hand an auditor a tablet and let them find what they need instead of sending the floor scrambling for a folder.
• It answers in any language, on any shift, so the same answer reaches every operator in their own words.
• Agents let you shape the process to how you actually work, whether that means five whys, a fishbone diagram, or a few extra steps that only make sense on your line.

The bigger shift is connection. Link your maintenance system, your quality system, scheduling, sensors, and your ERP, and an operator can make a deliberate call at the line instead of a knee jerk one pulled off a spreadsheet somebody built ten years ago. Bolt that onto a broken loop and it fails faster than the paper did. Build it on a loop that actually closes and it compounds.

Proof From the Floor

The two customer examples James walked through made the case better than any slide could.

USG, an early adopter of Poka’s AI-powered content tools, built 2,400 work instructions and saved thousands of hours. In the words of Alberto Macias, USG’s visual work instruction safety planner, “We created 2,400 instructions and saved thousands of hours. That speaks for itself.” For a continuous improvement team, that is the gap between capturing knowledge while it is still useful and watching it sit in someone’s head until they retire and take it with them.

L’Oreal was the other. New hires trained through Poka reached 11% higher OEE in their first three months, and as project manager Marc-André Lavoie put it, the team can track KPIs by category and by department to adapt its learning strategy over time.

What Phil zeroed in on was what USG really proved. The failure mode of most programs is building work instructions locked in an office. The value shows up when you build them at the Gemba, at the workstation, by the people who do the job. He has watched operators say we do not actually do it this way, this way is better, make the change on the fly, and push it through approval in hours instead of the days or weeks the old paper process demanded. The win was never digitizing content faster. It was capturing what the frontline already knows, where the work happens, and getting it shared before it goes stale.

The Part No Tool Can Do for You

James and Phil closed on the thing that gets lost in every AI conversation. The technology does not build the culture. Leadership commitment and everyday celebration do.

AI makes your people more valuable, not obsolete. It is a tool, worth exactly as much as the data, processes, and culture you already have, and no more. Try to make it a substitute for those things and it will fail. Use it to accelerate them and it becomes the next turn of a wheel that has been spinning for 140 years.

Watch the Full Session

The webinar goes deeper than any recap can, including the live demonstrations, the customer stories, and the audience questions on operator trust and who actually gets access to connected data. If you own continuous improvement at your plant, it is worth the hour.

Watch the full webinar recording

FAQs About Continuous Improvement and AI on the Factory Floor

What percentage of factory floor problems go unreported?

Around 68% of problems, defects, and improvement opportunities on the shop floor never get reported, usually because there is no simple way to report them and nothing happens when workers do raise an issue.

Why do continuous improvement programs fail?

They tend to break in three ways: digitizing paper without changing the habit underneath it, never closing the loop so gains are not sustained, and guessing at problems instead of acting on data, which fixes symptoms rather than root causes.

What does closing the loop mean in continuous improvement?

Closing the loop means capturing an issue, finding and documenting the root cause, putting a fix in place, preventing recurrence, and sustaining the gain with a standard so the floor does not slide back to old habits months later.

Does AI replace continuous improvement teams?

No. AI accelerates the existing rituals like root cause analysis and standard work rather than replacing them. It is only as effective as the data, processes, and culture already in place, and it makes frontline workers more valuable rather than obsolete.

How fast can manufacturers create digital work instructions with AI?

One Poka customer, USG, created 2,400 digital work instructions and saved thousands of hours using the AI-powered content tools built into the platform.