How Large-Scale Manufacturers Drive AI Value From The Factory Floor To The Boardroom

AI has arrived on the factory floor, and its nuts and bolts value is no longer theoretical. Falling compute costs and maturing data infrastructure are pushing large-scale manufacturing from reactive to predictive models, but the harder shift is organizational. Data governance, IT/OT convergence, and organizational trust now determine whether AI delivers or stalls.

Nick Giannakakis is the Group Chief Information Officer at Motor Oil and a board member at MORE Energy. He has led digital transformation at British American Tobacco and Coca-Cola Hellenic Bottling Company, and is a recognized Top 100 CIO and alumnus of executive programs at IMD, MIT Sloan, and London Business School. Giannakakis sees the industry undergoing a structural evolution, driven by durable economic and architectural changes.

"This is not hype or a bubble. We are seeing a true evolution," said Giannakakis. That foundation, however, doesn't build itself. Before any of those capabilities become operational, leaders in these environments face a connectivity and infrastructure challenge that is distinct from almost any other industry. In a manufacturing plant, that challenge starts at the most basic level: getting reliable data off the floor in the first place.

• The need for speed: "In manufacturing, establishing the right infrastructure is a challenge in itself. You need a closed LTE network, for example, just to collect the data points," he said. "That foundation is what enables Edge AI for situations where decisions in microseconds are imperative. A natural gas pipeline is a prime example of this." It is precisely these high-stakes environments that make the case for getting the infrastructure right from the start.

• Follow the star: With that groundwork in place, the destination becomes clearer. Giannakakis pointed to the digital twin as the industry's defining target, one that pulls together the key manufacturing trends now reshaping large-scale operations. "For us, the North Star is the digital twin. On top of a centralized information system, Generative AI now facilitates not just structural access to P&IDs, but also predictive risk modeling tied to our maintenance calendars. For the first time, we have these kinds of capabilities." Those capabilities, however, bring a new set of decisions about how much autonomy to hand over to the systems enabling them.

• Recommend or command?: These new tools introduce a core operational dilemma: should AI merely suggest actions, or be empowered to automate them? In practice, trust is earned incrementally, often by layering AI-driven decision-making on top of existing control systems. The journey is made possible by modern data fabrics that provide the infrastructure for new AI workflows. "In our industry, Advanced Process Control optimizers have been in place for years. AI now stands on top of them, using time-series models to provide recommendations. That is the first step," he noted. "As trust builds, those recommendations can be associated with direct action, where the system doesn't just propose but takes direct control."