Every manufacturer wants AI to transform operations. But two fundamental challenges are keeping the industry stuck in pilot mode—and neither of them is a lack of algorithms.
If you work in IT, you’ve spent the past decade watching AI reshape how software gets built, how customers get served, and how decisions get made. You’ve seen DevOps transform the relationship between development and operations. You’ve seen cloud-native architectures replace monoliths. You’ve seen organizations that moved fast pull away from those that didn’t.
Now imagine a world where almost none of that happened.
That’s manufacturing.
What We’re Actually Talking About
Manufacturing and industrial operations—think refineries, power plants, water treatment, automotive assembly lines—are the backbone of the physical economy. They are where raw materials become products, where energy gets generated, and where the laws of physics don’t come with an undo button.
The technology that runs these operations is called OT: Operational Technology. It includes PLCs (programmable logic controllers), SCADA systems (supervisory control and data acquisition), distributed control systems, historians that log decades of sensor data, and more. This is the tech that keeps turbines spinning, chemical reactions stable, and assembly lines moving.
Fifteen years ago, the term Industry 4.0 was coined, ushering in what was supposed to be the fourth industrial revolution. Fifteen years ago, a central control room of a refinery was among the most state-of-the-art workplaces you could find.
And yet, not that much has changed.
Of course, there’s been progress—IIoT deployments, some machine learning, better connectivity. But if we’re honest, most of it has been incremental. Our operators manage their entire personal lives from their phones and talk to a Large Language Model at home, but when they walk into the plant, it’s as if they’ve gone back to BlackBerry OS.
So what’s holding things back?
The Divide Between IT and OT
If you’ve lived through the DevOps transformation, you know what it means to bridge a gap between two technical groups with fundamentally different priorities. IT Development wanted speed and innovation. IT Operations wanted stability and reliability. The friction between them created slow deployment cycles, high failure rates, and a culture of blame.
DevOps emerged as a response. Not by mandating collaboration, but by engineering the context in which those teams worked. Different questions were asked: How do these teams actually interact? What information flows between them? Where do handovers fail?
We want to do the same thing—but for the gap between IT and OT.
And if you thought the gap between Dev and Ops was large, the one between IT and OT is the Grand Canyon.
The differences run deep. Different technical backgrounds. Different reporting lines—the CIO and the COO often don’t even share a strategy meeting. Different technology stacks—virtual versus physical. Different attitudes to change—IT thrives on low marginal costs and rapid iteration; OT has one foot permanently in the physical world, where adding a sensor means wiring, validation, and paperwork. And as is often the case, it’s the end users of that technology who are left stranded in the middle.
Why We Can’t Afford to Wait
We can’t spend another fifteen years making incremental improvements. The world has changed.
The pandemic showed everyone that manufacturing is strategic and essential to everyday life but also that our supply chains are fragile. Global superpowers now use manufacturing as a lever of power: Wars are waged with resources in mind, and rare earth minerals are used to exert geopolitical pressure. And the sustainability question looms over everything: How do we balance competitiveness with the imperative to not harm the planet?
On top of all that, cost pressures keep mounting. A silver tsunami of retirements is draining institutional knowledge—the experienced operators who can diagnose a problem by sound alone are walking out the door. And replacing them is getting harder every year.
If we can bridge the IT/OT divide, we can start using technology to help solve these problems. If we can’t, we’ll keep building smarter silos instead of smarter plants.
Where AI Sits Today
In our ETLS talk last September, we introduced a framework for understanding how AI fits into industrial operations. We walked through the physical twin (the actual plant), the digital twin(s) (the many systems that represent it digitally), the automation layer, and the concept of a Virtual Operator.
Today, AI in industrial settings is essentially a bolt-on onto those digital twins.
A very clever bolt-on in some cases, but a bolt-on nonetheless.
On the digital twin side, AI tools help you search through sensor data faster, spot patterns in quality data, and optimize production planning. On the automation layer, it can act as a coding assistant, help configure systems, or aid operators in analyzing events.
Useful? Absolutely.
A step change? Not yet.
Because the fundamental relationship between the digital and the physical world hasn’t changed. Operators and engineers still bridge that gap manually. Automation still handles the deterministic stuff—the processes where inputs reliably lead to the same outputs. And for everything outside those boundaries—the unforeseen, the non-deterministic—humans step in.
So what needs to change to move beyond bolt-ons?
Three Steps Towards the Virtual Operator
To frame that question, it helps to think in three stages of autonomy.
- Step 1—Assistance (where most of us are today). AI acts as a digital assistant. It gets access to a portion of your data and helps you get work done more effectively. Think of it as a smarter search engine for your plant.
- Step 2—Collaboration. AI becomes a virtual co-worker. It has enough context to recommend actions, flag anomalies, and support decision-making. But it has no direct control of the plant. It suggests; it doesn’t act. This is the equivalent of lane assist or adaptive cruise control in your car.
- Step 3—Autonomy. The AI system can independently handle certain operational decisions. In most cases, this requires a fundamental rethink of your plant’s control and safety concepts. And we’re not there yet—not by a long shot.
The self-driving car analogy is useful here. The hard part of autonomous driving isn’t controlling the steering wheel. It’s the fact that every intersection is different, road conditions change constantly, and there will always be unforeseen circumstances. Industrial operations face the exact same challenge. Equipment degrades in unpredictable ways. Raw materials vary batch to batch. Ambient conditions affect process behavior. The margin for error in a chemical plant is rather different from the margin for error in a software deployment.
Challenge #1: The Integrated Digital Twin
To move from assistance to collaboration, AI needs integrated data and context. Not just access to one historian or one maintenance system, but a coherent view across all your digital twins.
Today, those systems are typically standalone. Even when they are connected, they use different naming conventions, different data models, and different levels of detail. Linking them in a unified, structured, contextualized way is genuinely hard work.
If you come from the IT world, think of it this way: imagine trying to build a recommendation engine where your customer data, transaction history, product catalog, and support tickets all live in different databases with different schemas, different keys, and no shared identifiers. Now multiply that complexity by the fact that some of these systems run on protocols from the 1990s and can’t be taken offline for migration because they’re controlling live processes.
An integrated data layer is a prerequisite for everything we want to do with AI at scale. Without it, you’re optimizing locally. You’re making a smarter silo, not a smarter plant.
The technology to build this layer exists. The complexity isn’t purely technological any longer—it’s the sheer volume of work it takes to integrate systems, align data models, and get your data governance right. And let’s be honest: In most cases, you need to build that model from scratch. It won’t magically drop out of the air.
Challenge #2: Understanding the Physical Twin
Here’s the harder problem—and the one that doesn’t get talked about enough.
Even if you solve the data integration challenge, your virtual operator still only sees the digital world.
For a virtual operator to truly collaborate—let alone act autonomously—it needs to understand what’s actually happening in the physical plant: the context, the constraints, the unwritten rules, the things your experienced operators just know but have never documented.
Some startups are beginning to tackle this by parsing engineering diagrams, ingesting technical literature, and building physics-informed models. That’s promising work. But this is an extraordinarily tough challenge, and we’ll be working on it for years—probably decades.
Here’s the uncomfortable truth: It is simply not possible to use historical data alone to understand the physical twin. We’ve seen companies try it, and they fail. You cannot infer physical reality purely from its digital shadow. Full stop.
And as long as we can’t bridge that gap, there is no replacement for the human operator.
Is That Pessimistic?
We don’t think so. We think it’s realistic.
The progress in industrial AI is real. The assistance phase is delivering genuine value, and the tools are getting better every month. But when you see vendors claiming they’ll deliver autonomous operations with a bit of historical data and a large language model…it’s worth pausing and thinking through the fundamentals.
Taking things step by step isn’t pessimism. It’s how you build something that actually works—safely and sustainably.
The data foundation is everything. Start there. Get your integrated digital twin in order. And keep an eye on the physical twin challenge—because that’s where the real breakthroughs will eventually come from.
Go Deeper
This article is part of our Industrial AI Unpacked series at the IT/OT Insider, where we cut through the hype and focus on what actually works.
If you want to see the full framework in action, David walks through it in this video:
📺 Watch: The Virtual Operator Framework
We also presented this thinking at the Enterprise Technology Leadership Summit (ETLS). You can watch the full talk here:
If you’d like to go even deeper—into data platform architecture, IT/OT cooperation models, and the practical frameworks behind all of this—check out ITOT.Academy, our live, vendor-neutral learning program built for IT and OT professionals who want to make collaboration work.
And if you want the full story—from the Purdue Model to cooperation patterns to building an industrial data platform—we’ve put everything we’ve learned into The IT/OT Handbook, coming this October from IT Revolution Press.
Willem van Lammeren and David Ariens are the cofounders of IT/OT Insider, a content and education platform focused on bridging the gap between IT and OT in manufacturing, process industry, and critical infrastructure.