By: GJD
When you’ve spent enough years around business systems, you start to notice a pattern. Technologies arrive with great fanfare, promise to change everything, and then quietly settle into whatever role the business actually needs them to play. Some fade quickly. Others linger far longer than anyone expected. And a few—usually the ones built with a clear understanding of how work really gets done—become the quiet backbone of entire industries.
My own career began in the era when MRP was still finding its footing. The systems were simple by today’s standards, but they were grounded in the operational logic that Oliver Wight, George Plossl, and Joseph Orlicki had been teaching for years. Those early systems didn’t try to be everything. They tried to be right. And because they were built around the realities of planning, scheduling, and control, they earned a level of trust that newer systems often struggle to match.
Over the decades, I watched wave after wave of technology arrive—each one promising to replace what came before. Some did. Many didn’t. But through all of it, COBOL systems kept doing their work in the background, quietly and reliably. There are an estimated 220 billion lines of COBOL still in active use across business sectors like banking and finance, insurance, government and transportation. For a long time, I assumed they were living on borrowed time. Then AI arrived, and I found myself rethinking assumptions I had carried for most of my career.
Why COBOL never really left
There’s a misconception that COBOL survived out of stubbornness, as if companies simply refused to modernize. That’s not what I saw. The organizations that kept their COBOL systems did so because those systems worked. They were stable, predictable, and deeply aligned with the business processes they supported. They had been refined over years—sometimes decades—of real-world use. And they were fully paid for, which is not a trivial detail when you’re responsible for a budget.
The real challenge wasn’t the technology. It was the people. As the generation that built and maintained these systems moved into retirement, the talent pool shrank. Companies weren’t afraid of COBOL; they were afraid of being unable to support it. That fear drove many modernization efforts—not dissatisfaction with the systems themselves.
I’ve seen what happens when organizations try to replace a system that has been quietly doing its job for thirty years. Users who were confident and productive suddenly find themselves wrestling with unfamiliar workflows. Productivity dips. Morale suffers. And sometimes, after enormous investment, the new system simply doesn’t work as well as the old one. Ford’s Everest project is one example among many: a massive effort to replace a legacy system that ultimately had to be abandoned, leaving the original system in place.
These experiences leave a mark. They teach you that “modern” is not always synonymous with “better,” and that the logic embedded in those older systems often reflects a depth of operational understanding that newer platforms struggle to replicate.
How modern systems drifted from their roots
As ERP vendors expanded their ambitions, the systems grew larger and more abstract. They tried to cover entire industries end to end, which meant they had to generalize processes that were once tailored to specific operational realities. Somewhere along the way, the focus shifted from supporting the business to shaping it. Companies found themselves adapting to the software rather than the other way around.
The result was complexity—layers of configuration, customization, and integration that made implementations long, expensive, and fragile. These systems certainly brought new capabilities, but they also introduced new risks. And they often lacked the clarity and discipline that defined the early MRP and ERP designs.
Meanwhile, the legacy systems kept running. They weren’t glamorous, but they were aligned with the business in a way that newer systems sometimes weren’t. They had become, in a very real sense, the operational memory of the organization.
Legacy systems as strategic assets
One of the things that becomes clear when you’ve lived with these systems long enough is that they contain more than code. They contain decisions—thousands of them—made by people who understood the business intimately. They contain refinements that were added over years of real-world experience. They contain data that has been collected consistently and cleanly for decades.
These systems were built with a kind of craftsmanship that is easy to overlook. COBOL handled the business logic, while assembly routines handled the performance-critical operations. It was a pragmatic architecture, tuned to the hardware and the needs of the business. For years, that architecture was seen as a barrier to modernization. Today, AI can analyze both layers, understand their interactions, and help migrate or extend them without losing the qualities that made them successful.
When you replace a system like that, you’re not just swapping out software. You’re risking the loss of institutional knowledge that has been encoded over decades. That’s why so many replacement projects struggle. The legacy system wasn’t just a tool; it was a reflection of how the business actually worked.
AI changes the equation
What AI brings to this landscape is not a new programming language or a new platform. It brings understanding. It can read COBOL code, explain it, document it, and refactor it. It can analyze assembly routines that once required a specialist to decipher. It can help organizations modernize selectively—extending what works, replacing what doesn’t, and wrapping the core with new capabilities.
For the first time in decades, companies are no longer forced into a corner. They don’t have to choose between keeping a system no one can maintain or replacing it with something that may not fit the business. AI creates a third option: continue with confidence. Modernize without disruption. Build on the foundation rather than tearing it out.
This is not a step backward. It’s a recognition that the systems built decades ago were grounded in principles that still matter—clarity, determinism, alignment with real-world processes. AI doesn’t erase those principles. It amplifies them.
A shift in how we think about modernization
As AI takes over the mechanical work of understanding and maintaining legacy systems, the role of human expertise shifts. The consultants who once specialized in specific platforms will find that their advantage is diminishing. What becomes valuable instead is the ability to understand the business itself—its processes, its constraints, its culture.
In a way, this brings us back to the roots of the field. The early pioneers weren’t system technicians. They were business thinkers. They understood operations, not just software. AI pushes us back toward that model, where the most valuable skill is not knowing how a system works, but knowing how the business works.
Modernization becomes less about replacing systems and more about improving them. Less about disruption and more about continuity. Less about technology for its own sake and more about aligning technology with the realities of the business.
Looking ahead
Over the next few years, I expect to see more organizations choosing to stabilize and extend their legacy systems rather than replace them. Not because they’re afraid of change, but because they recognize the value of what they already have. AI gives them the ability to modernize without losing the logic, the data, and the operational alignment that have served them well.
The future will not be monolithic. It will be hybrid—stable COBOL cores augmented by AI-driven intelligence, automation, and analytics. It will be a future where reliability and innovation coexist, where modernization is incremental rather than traumatic, and where organizations gain the benefits of new technology without sacrificing the foundations that have kept them running.
COBOL isn’t returning. It never left. What’s returning is our appreciation for the systems that were built with care, grounded in sound principles, and refined through decades of real-world use. AI simply gives us the tools to carry those systems forward.
