To get value from AI, first define your business in nouns and verbs
AI has moved from experimental to expected in manufacturing, but results haven’t kept pace with ambition. In fact, 95 percent of generative AI pilots fail.
One reason: starting in the wrong place, with the wrong data.
AI can’t optimize what your organization hasn’t clearly defined. To succeed, it needs a domain model—a breakdown of the nouns and verbs of your business: core entities, key relationships, constraints, and actions (basically, your tables and APIs)—before it can carry out predictive maintenance, smart routing, and automated optimization.
Here, I'll explain how a domain model is the missing link between AI potential and meaningful ROI in manufacturing contexts.
Domain models provide the baseline your AI needs to succeed
Most manufacturing systems are backward-looking. Reports focus on what’s already happened. Data lives in silos. Asset health is evaluated only after a failure.
In other words, assets are monitored but not modeled, so they can’t be intelligently managed in real time.
So what’s the solution?
To turn data into intelligence, you must first define the language of the asset ecosystem through structured tables and APIs—that’s where domain modeling comes in.
Domain modeling asks deeper questions about your business that give AI a baseline from which to operate:
What are the core entities (machine, component, work order, technician, line, material, etc.)?
- What relationships exist?
Which components affect which machines?
What defines a bottleneck?
What constitutes downtime?
What are the actions (inspect, repair, reroute, schedule, prioritize, etc.)?
Before it can predict failures, your AI must know what failure means. Before it can recommend preventive actions to preserve asset health, it must know how you define asset health—and which trade-offs matter (cost vs. output vs. quality).
Without that clarity, your AI is optimizing against vague objectives that don’t ladder up to success for your organization.
A strategic shift: from asset chaos to intelligent orchestration
TXI works with a railcar repair company juggling many challenges at once: a high volume of work, reactive scheduling, loosely tracked assets, and limited visibility into bottlenecks.
Our first step to help them was to build a model that defined their business. It described railcar, job, and shop. We built structured tables and APIs that represented the asset relationships, work flows, and constraints.
At first, the work felt abstract. That’s normal. But once the model existed, the company could apply specific strategies. Executives coded asset prioritization, systems dynamically allocated resources, and machines responded in real time.
With this groundwork laid, the company’s AI had the vocabulary to understand what was happening and the know-how to intelligently manage assets in alignment with business strategy.
That’s just one example of how a domain model enables intelligence. Having a model lets you go from traditional asset management—in which your system monitors assets and reports a failure, so it can be fixed—to AI-enabled asset orchestration.
When you’ve made asset orchestration possible, AI can detect patterns humans can’t see fast enough to act on. It can anticipate failures and prioritize preventative maintenance during downtime. It can reroute work when one machine shows signs of failure.
Beyond your factory floor, a system like this can factor in external signals—when a material cost spikes, for example, it might recommend shifting production from the affected products to another until the material price stabilizes again.
A domain model turns asset strategy into executable logic. It means that AI is immediately knowledgeable about your business. Without the constraint of a learning curve, automation can move at the speed of the machine.
Misconceptions about AI-powered asset management
When I’m helping companies implement successful AI solutions, manufacturing leaders often bring up these two misconceptions that hold back progress on their AI strategy.
1. The flashiest AI tool is the best AI tool.
Tool-first thinking is why most AI initiatives stall after the pilot phase. Some manufacturing leaders get excited by a new AI tool and try to make it fit their business.
They apply AI at the surface level only, adding sensors and dashboards that are disconnected from a domain model. Data is fragmented, recommendations are inconsistent, and trust in the AI erodes among employees. It’s like trying to fix your car with a paint job when the engine’s on the fritz.
The developer’s adage “garbage in, garbage out” applies here. AI amplifies your system’s existing foundation. If that foundation is vague, AI’s output will be too.
2. AI will replace our people.
This is a common concern that holds leaders back from fully investing in AI. And while AI is a powerful tool, it’s certainly not capable of replacing valuable human experience. Think of AI more like a co-pilot for your machine operators, working in tandem with them.
Another way to think about the AI-human advantage is by comparing it to Grammarly, the software you might use to help draft emails or suggest improvements to your writing.
It provides recommendations, flags mistakes (risks) in real time, and suggests optimal responses—but Grammarly doesn’t force you to accept its suggestions. Similarly, your operators can choose to accept, decline, or refine AI’s suggestions.
AI won’t replace your asset managers. It will help scale their judgement.
Context first, scale second
By representing asset workflows, relationships, and constraints, a domain model provides the foundation your AI needs to be successful, but that doesn’t mean you should try to solve every problem at once.
Take a pragmatic approach. Identify a contained problem with measurable financial impact and set ambitious KPIs—AI is there to help supercharge your workers’ performance, after all. Use the results to start building a business case for why AI-powered asset management is worthwhile on a larger scale.
Imagine one machine in your factory breaks down monthly. The downtime from each failure costs $10,000 in lost revenue. You use AI to manage the machine and, ultimately, prevent its breakdowns.
In a year, you’ve saved $120,000 on just that machine. By calculating your savings, you start to see the impact that AI-powered asset management can have across your entire factory.
None of this works unless your AI understands the context it’s operating in. Before you buy another tool, define your business in nouns and verbs. Success builds from there.
About the author
Kyle McCluskey is a Lead AI Solutions Engineer here at TXI. Kyle bridges the gap between AI technology and real business outcomes, working alongside client teams to build scalable, production-ready data systems that go beyond pilots. With a background in computational linguistics, a master's degree in AI, and over a decade launching data engineering teams, Kyle brings both technical depth and practical judgment to every engagement.
