In my last article, I introduced a simple but unforgiving equation:
Impact = Technology × Behavior × Management.
If any of the three components approaches zero, the product collapses. Impressive tools, great people, or strong leadership — on their own — are never enough. Only when the three interact do we see lasting improvement.
In this next set of articles, I’d like to unpack each part of that equation a little further. Let’s start with the piece that gets all the current attention and funding: technology.
As lean practitioners, we already know something uncomfortable: new technology almost never succeeds due to its implementation. It fails in the gap between what it could enable (i.e., the compelling reason why you chose to deploy it) and how it gets used.
We’ve seen this before — on the shop floor, in the office, and now with everything touched by AI.
The Andon Board That Didn’t Change Anything
If you’ve worked in operations long enough, you’ve seen some version of this story.
A facility decides to “become more like Toyota,” so it invests in an andon system: lights, boards, maybe even some nice screens with real-time metrics. The launch event is enthusiastic. Photos are taken. A few visitors are even impressed.
Then … not much happens.
- Operators don’t pull the cord when they see a problem because nobody responds quickly — or they get subtle pushback for “slowing things down.”
- Team leaders are busy firefighting and treat the signal as a nuisance rather than a work priority to respond to a team member.
- No one reviews the andon history in a daily or weekly PDCA cadence to see patterns and remove root causes.
Technically, the system works. The lights turn on. The board updates. The dashboard looks modern.
But operationally, nothing has changed. The andon has become what I once called lean wallpaper — nice to look at, but irrelevant to performance.
From a Technology × Behavior × Management standpoint, you can see why:
- Technology: The andon system is fine.
- Behavior: People don’t consistently signal and respond in the intended way.
- Management: Leaders haven’t built andon into daily routines, escalation paths, or problem-solving expectations.
Multiply those together and you get roughly zero impact.
We’ve Seen the Same Movie in the Office
Move from the shop floor to the office and the pattern repeats.
Think back to large IT programs over the past 25 years:
- Y2K remediation projects that upgraded systems but never addressed broken processes.
- ERP implementations that standardized transactions but didn’t clarify ownership, decision rights, or problem-solving responsibilities.
- Workflow tools and digital dashboards that made queues more visible but didn’t change how leaders prioritized or removed bottlenecks.
In each case, the promise was similar to what we hear around AI today: “This system will give us real-time visibility, better decision-making, and a single source of truth.”
Sometimes that happened. Often it did not.
Why? Because technology alone is potential energy. Until it is connected to new behaviors and management systems, it just sits there.
When organizations focus on tools alone, they typically end up with scattered pilots, uneven adoption, and little measurable impact. That is one of most important learning points from working at Toyota in Japan. We were very thoughtful about technological implementation and how to integrate it with people and management.
AI Is Following the Same Curve
Now a new wave is breaking. What I call “narrow AI,” such as generative AI, copilots, agents, and automation is now exploding. If you listen to the hype, you might think this time is different. But from a lean perspective, the adoption pattern looks very familiar.
I see many organizations racing to:
- Launch AI initiatives or digital tools.
- Roll out generic copilots to thousands of employees and tell them to “use” them.
- Announce ambitious automation targets.
Yet when you go to the gemba — whether that’s an engineering floor, a call center, or a hospital — it’s not always clear what problem AI is actually solving.
From the outside, it can look like AI is magical in one company and useless in another. Under the hood, the difference is found in the value associated with each component: Technology × Behavior × Management.
Technology Is the Easy Part
This is a strange statement coming from someone who once struggled to write a simple API call in JavaScript or Python. But after a couple years of working with the large language models (LLMs), I’ve come to believe the following:
In this era, technology is often the easiest variable to change.
- The cost of computing continues to drop relative to capability.
- Powerful models are available via API with a credit card and a bit of documentation.
- Open-source tools and community examples shorten the path from idea to prototype.
By contrast, changing how people work and how leaders manage remains hard, slow, and context-specific.
That’s why many AI initiatives, I predict, will quietly underperform. It’s not that the model isn’t powerful enough. It’s that the implementation lives entirely on the left side of the equation:
Technology × (Low Behavior) × (Weak Management) ≈ Minimal Impact
The result feels a lot like that andon cord nobody pulls.
What Technology Can Do (When Connected Correctly)
None of this is an argument against technology. Quite the opposite. When it is tightly linked to behavior and management, even modest tools can deliver outsized results. I’ve seen simple, well-integrated solutions do remarkable things:
- A production board and 10-minute standup meeting, when used daily, transforms communication, surfaces problems earlier, and accelerates kaizen.
- A narrow AI tool can review problem-solving reports, catch logic gaps, and act as an “in-between coach” when human experts aren’t available.
- A small assistant using retrieval-augmented generation (RAG) helps engineers find standards, past countermeasures, and lessons learned in seconds instead of hours.
- A vision system with machine learning catches defects in raw materials and work-in-process that was not possible before.
In all those cases, the enabling technology was important — but not sufficient. The real leverage came when leaders answered questions like:
- Who uses this, at what moment, in which workflow?
- What problem is it supposed to help us see or solve more clearly?
- How will we check whether it’s actually changing outcomes?
The tools became part of the system of work, not just another app on the corporate tool menu.
Four Questions for Leaders Before You Buy the Next Tech Tool
In the next two articles, I’ll dig into behaviors and management systems in more detail. For now, I want to leave you with four practical questions aimed squarely at the technology side of the equation.
These questions won’t tell you which model to buy or which vendor to trust. They will help you avoid turning your next investment into digital wallpaper.
1. What specific problem is this technology supposed to help us solve and how?
If you can’t state this clearly in one sentence, you’re probably starting from the tool, not from value. “Using AI” is not a problem statement. It is an action item disguised as a problem statement.
2. Exactly where in the workflow will it live — and for whom?
Name the role, the moment, and the trigger. “Supervisors will use this AI assistant for pre-shift huddle prep when they review yesterday’s abnormalities.” If you can’t be this concrete, the tool will float above the work rather than inside it.
3. What observable behaviors should change when the technology is working?
For andon, it might be “Operators pull the cord earlier and leaders respond within 60 seconds.” For AI, it might be “Employees check their problem-solving logic with the assistant before actual meetings or reviews.” If you can’t describe the behavior change, you won’t be able to coach or measure it.
4. How will this technology plug into our existing management system?
Where will its results show up — in daily huddles, weekly reviews, A3 discussions, hoshin catchball? Who will look at the data and ask, “What are we improving and how?” If the answer is “I don’t really know,” impact will be accidental at best.
You’ll notice that all four questions start with technology and immediately move into behavior and management. That’s intentional. In practice, you can’t completely separate the three. But you can discipline yourself and not stop after the selection and deployment of the tool.
Closing Thought
From andon boards to ERP systems to today’s AI copilots, we’ve been here before. The logos on the hardware and software change; the underlying pattern does not.
The organizations that will benefit most from AI are not the ones with the biggest models or the flashiest demos. They are the ones that treat technology as one factor in a larger equation — connecting it deliberately to how people work and how leaders manage.
In the next article, I’ll turn to the second term in that equation: behavior — the skills, habits, and culture that determine whether any tool becomes part of real problem solving or just another icon on the screen.
For now, remember: technology is necessary but never sufficient. Neither the latest AI model nor the andon board alone changes anything — people and systems do.
Get insights into AI applications through the lens of Lean Thinking delivered to yout inbox. Subscribe to the Lean AI Journal on LinkedIn.
2026 Lean Summit
The premier leadership conference shaping the future of lean management for every business.
