AI is changing manufacturing in ways that would have been difficult to imagine even a few years ago. Companies are using it to analyze production data, streamline documentation, identify maintenance trends, and make technical information easier to access.

As experienced engineers, technicians, and maintenance specialists retire, AI will become an increasingly valuable tool for preserving decades of technical knowledge that might otherwise disappear.

See also: Supporting the critical role of system integrators

But after spending nearly three decades working with aging infrastructure, I've found myself thinking about a different question: Where does the knowledge we want AI to preserve actually come from?

It's an important distinction because AI doesn't create institutional knowledge, it organizes it. Every maintenance record, inspection report, SOP, and best practice exists because someone encountered a problem, solved it, documented the solution, and refined it over time. Before knowledge becomes searchable, it begins as experience.

While my experience comes from a specialized corner of manufacturing, every manufacturer depends on knowledge that has accumulated gradually through production, maintenance, quality control, and problem-solving.

Engineering drawings, operating manuals, and technical specifications provide an essential foundation, but they only tell part of the story. The rest comes from years of working with products under real operating conditions, where unexpected challenges force people to adapt, improve, and document what they've learned.

As manufacturers invest in AI to organize and preserve what they know, it's worth remembering that technology can only build on knowledge people have already created.

Experience creates knowledge long before it's documented

One of the biggest misconceptions about manufacturing expertise is that it begins with innovation. In reality, most expertise develops much more gradually.

Processes improve because operators refine them over hundreds of production runs. Maintenance procedures evolve because technicians repeatedly encounter the same failure and discover more reliable solutions.

See also: Understanding connected worker solutions: Expectations vs. reality

Quality standards become more robust because inspectors notice subtle patterns that weren't obvious during earlier production cycles. Those incremental improvements over time become accepted practice, even though few people can identify the exact moment they became standard.

My own experience followed that same pattern. Early restoration work involved constant experimentation. We tested different approaches, observed how materials behaved after years of environmental exposure, documented what worked, and adjusted our methods on the next project.

Individual lessons slowly evolved into repeatable procedures, inspection methods, and quality standards. There was no single breakthrough that transformed the work; only hundreds of small improvements accumulated over time.

That's true across manufacturing. Continuous improvement rarely comes from one dramatic discovery. More often, it develops through thousands of practical decisions made by people working closest to the product.

What products teach after they leave the factory

Manufacturers invest enormous effort designing, testing, and validating products before they enter service. Engineering drawings define dimensions and tolerances. Material specifications establish performance requirements. Production testing verifies that products meet quality standards before they leave the factory.

But products continue teaching us long after manufacturing ends.

Once equipment enters service, it begins operating under conditions that no laboratory or prototype can fully reproduce. Different maintenance practices, operating environments, weather conditions, workloads, and repair histories gradually shape how products perform over years or decades. Those experiences create a second layer of knowledge that often receives far less attention than the original design process.

See also: AI has a trust—not a technology—problem

That's something I came to appreciate after years of working with equipment that had already accumulated decades of service. The biggest challenges rarely involved understanding how a product was originally designed. They involved understanding everything that had happened since it left the factory.

Equipment that looked nearly identical on paper often aged very differently in practice. Components expected to require frequent attention sometimes proved remarkably durable, while seemingly minor design decisions occasionally became recurring maintenance challenges.

After seeing enough examples, patterns begin to emerge. You stop looking only at the repair in front of you and start recognizing the decisions that shaped the product's entire lifecycle.

That perspective doesn't replace engineering. It strengthens it.

Manufacturers already rely on laboratory testing, warranty data, and customer feedback to improve future products. Field experience deserves a place alongside those sources because it reveals how products actually perform after years of operation.

In many cases, the people best positioned to recognize those lessons aren't product designers. They're maintenance teams, inspectors, operators, and technicians responsible for keeping equipment running safely every day.

Some of the most valuable insights I've gained throughout my career came from those conversations. Experienced mechanics and maintenance professionals understand products differently because they live with them throughout their operational lives.

They know which repairs continue performing years later, which components consistently require attention, and which design features quietly create unnecessary work every maintenance cycle.

See also: Physical AI: Where opportunity really sits—now, today—for manufacturers

Too often, that knowledge remains informal. It gets shared between experienced employees but never becomes part of an organization's permanent knowledge base. As more experienced workers retire, manufacturers risk losing decades of practical insight that could improve future products, simplify maintenance, and strengthen long-term reliability.

AI can preserve expertise. It can't create it

This is where I believe AI has the greatest opportunity; not as a replacement for experienced manufacturing professionals, but as a way to preserve and scale the knowledge they've already created.

Manufacturing is entering a period of significant workforce transition. Experienced engineers, machinists, inspectors, technicians, and maintenance specialists are retiring across the industry, taking with them years of practical knowledge that often exists nowhere else.

Sometimes it's understanding why a process changed years ago. Sometimes it's knowing which repair consistently outperformed the alternatives or recognizing the early signs of a recurring problem before it becomes a costly failure.

Much of that expertise never makes its way into formal documentation.

AI has the potential to change that. Manufacturers can use it to organize decades of maintenance records, inspection reports, engineering change notices, and technical documentation into knowledge that's searchable, connected, and far easier to access. Instead of relying on individual memory, organizations can begin building systems that preserve not only technical information, but also the experience behind it.

See also: DX is diminished if your workforce isn’t upskilled

That could shorten training, improve consistency across teams, and help younger employees build on what previous generations have already learned instead of rediscovering the same lessons themselves.

At the same time, it's important to recognize the limits of what AI can do. It can identify patterns in existing information, but it can't generate the experience that created that information in the first place.

Throughout my career, I've occasionally worked with legacy equipment whose original tooling, replacement parts, or technical documentation no longer existed. Solving those challenges required more than retrieving historical records.

It required understanding how products had actually performed after decades of service, why previous repairs succeeded or failed, and how modern solutions could improve long-term reliability while remaining compatible with existing systems.

Experiences like that reinforced something I believe applies across manufacturing: products continue generating valuable engineering knowledge long after production ends, but only if organizations capture those lessons before they're lost.

Rather than asking whether AI will replace manufacturing expertise, I believe the more useful question is how it can help manufacturers preserve, organize, and expand it.

Treat institutional knowledge like a manufacturing asset

Manufacturers have long understood the importance of investing in equipment, automation, and process improvements. Institutional knowledge deserves the same attention.

That starts by recognizing that valuable expertise exists throughout an organization, not only within engineering or leadership teams.

Maintenance technicians, operators, inspectors, machinists, and production employees often see problems, and opportunities for improvement, long before they become formal engineering discussions because they're working with products every day throughout their operational lives.

Capturing that knowledge requires recording why processes changed, encouraging experienced employees to mentor newer team members, and creating feedback loops that allow lessons learned in production and maintenance to influence future product development. The organizations that do this well don't simply preserve knowledge, they continuously improve it.

See also: The fiber bottleneck nobody priced into the AI boom

AI can make those efforts far more effective by connecting information that would otherwise remain isolated across departments, facilities, or generations of employees. But technology works best when it's supporting a culture that already values learning, documentation, and knowledge sharing.

Institutional knowledge isn't static. Every production run, maintenance cycle, customer issue, and engineering improvement adds another piece to an organization's understanding of how its products perform in the real world.

The companies that continue improving over decades are usually the ones that treat those experiences as valuable data rather than isolated events.

AI will undoubtedly become an important part of manufacturing's future. But I believe its greatest contribution won't be replacing human expertise. It will be helping manufacturers preserve decades of practical knowledge while giving the next generation a stronger foundation to build upon.

Because the most valuable manufacturing knowledge is created by people who observe carefully, solve difficult problems, learn from experience, and ensure those lessons don't disappear with them.