Council Post: Creating A Digital Backbone For AI-Driven Manufacturing

Aron Semle is the Chief Technology Officer at HighByte.

getty

When I was eighteen, I worked on the night shift at a Poland Springs bottling facility in Hollis, Maine. One evening, as my supervisors ran the facility at 1.25x speed to make up for a shipment delay, we experienced what every manufacturer dreads: a jammed conveyor belt.

We responded as any other team would: shutting down the malfunctioning line, investigating the issue and searching for a quick workaround. With no time to lose, we chose to manually shift the bottles to another nearby conveyor, assuming that it was open and available. Once the bottles were moved, we fired up the line.

Seconds later, a technician came flying out of the next room, cursing at us. He’d been in the middle of servicing this alternative line, and we’d unknowingly put both him and the machinery at risk with our actions. Once again, operations came to an immediate stop, and we were right back where we started—only now we’d wasted time, effort and dug ourselves into a deeper hole than before.

Now, if I asked you about the root cause of this series of unfortunate events, you’d likely say it was the jammed conveyor belt. But I’d argue that the main issue, the one that still plagues manufacturers to this day, is something much larger and more systemic to the industry: facility-wide information disconnect.

The Illusion Of Available Data

What happened at the bottling plant wasn’t due to a lack of available information. Operational data is nothing if not plentiful in modern manufacturing environments, generated and stored across machines, PLCs, HMIs, CMMS, ERP systems and MES applications. The issue is that this data is not connected and reviewed in the broader context of the plant.

Today’s industrial environments might look connected, but from an IT perspective, they are very much not. Legacy architecture creates data silos, with systems that only communicate through narrow, vertical pathways and point-to-point integrations.

When every system requires custom connections, you end up with fragile, hard-to-maintain pipelines in which data is duplicated or inconsistent. It exists, but it’s not discoverable, standardized or contextualized in ways that make it useful to plant employees or AI models. It lacks any shared naming conventions or broader alignment on the business context.

This fragmentation is what inhibits effective data-driven decision-making on the factory floor. Teams are uninformed, connected solutions cannot scale and decisions are often delayed or reactive—or, in the bottling example, potentially dangerous. This challenge will not be solved by just creating better integrations between disparate systems, but by adopting a context-driven architectural foundation.

Empowering The Factory With Context

Regardless of the solutions involved, today’s most effective manufacturing architectures share several characteristics:

• They provide a common way to organize data around equipment, processes and business assets.

• They make information accessible across OT and IT teams and systems.

• They allow new systems to consume data without requiring another set of custom integrations.

While evaluating ways to modernize their data infrastructure, manufacturers should prioritize models that:

• Create shared context across operational systems.

• Support real-time access to data.

• Reduce dependency on point-to-point integrations.

• Scale across sites and business units without requiring extensive customization.

• Enable both human users and future technologies to consume data consistently.

Ultimately, this layer should enable teams to stop asking “Which system holds this data?” and start asking “What’s actually happening on this machine, in this process, right now?” It should move architecture from system-centric to context-centric, creating a unified view of manufacturing operations. When paired with a clear data strategy, manufacturers achieve much more effective alignment among operations, engineering, IT and shop-floor teams.

​From Contextualized Data To Intelligent Operations

Creating a shared data layer for systems to exchange information within a common structure and context helps create a reliable view of operations across the plant, allowing users to make more informed, data-driven decisions. In the bottling scenario, the goal would not be to simply connect the maintenance and production systems, but to create a shared view where maintenance activity, equipment status, and production data could be understood within the same context.

Empowered by shared conventions, our team could evaluate coherent data describing what’s actually happening on the floor—not a smattering of different terms and measurements from different systems. Fragmented signals, like those from a broken conveyor belt and a line under maintenance, could be turned into a clear, unified signal, and we could pivot to another plan without running into more trouble.​

The Importance Of Contextualized Data

What makes these digital backbones particularly timely is their relationship to industrial artificial intelligence. Industrial AI applications aren’t limited by the capabilities of their algorithms; they simply don’t have access to enough connected data. Unified data layers help provide the real-time, structured and contextualized data that these solutions need to deliver value.

The most direct application of AI across today’s industrial sector is in combating the impending “silver tsunami.” As factory experts age towards retirement, they’re taking decades of institutional knowledge with them, creating significant skills gaps and increasing the risks associated with inexperience and inefficiencies.

Truly Connected Digital Manufacturing

Whether in a bottling plant, an oil and gas refinery or an automotive factory, the industrial sector’s biggest data challenge is the inability to connect, contextualize and act upon the data that’s already being generated.

By creating a digital backbone that transforms fragmented information into a coherent, real-time, accessible view of operations, teams can empower operators and AI systems to make better, more informed decisions in any scenario—even in the midst of a night-shift production crisis.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?