The expertise that keeps industrial plants running efficiently often lives in the minds of experienced operators. As those operators retire, that knowledge is at risk of being lost permanently. AI offers one of the most practical tools available for capturing and preserving it.
In the control rooms and on the process floors of the world’s most carbon-intensive industries, a form of expertise exists that does not appear in any manual. It is accumulated over decades: the instinct a furnace operator develops for how a heat behaves under particular atmospheric conditions; the adjustment a kiln supervisor makes before any instrument signals that something needs attention.
This is what engineers call tribal knowledge. And a substantial amount of it is currently leaving the industry.
The Demographic Shift Facing Heavy Industry
Heavy industry is facing a significant generational transition. Experienced operators who entered the sector in the 1970s, 1980s and 1990s are retiring in large numbers. The expertise they carry was never systematically documented. It developed through years of hands-on interaction with physical systems and was refined through accumulated experience of what works, often through failures that younger operators have never encountered.
The consequences are significant for both safety and efficiency. Plants that have operated reliably for decades can become less stable when the people who understand their particular characteristics depart. Less stable operation means less efficient operation. Less efficient operation means more energy consumed and more emissions produced per unit of output.
The IFS 2025 global study The Invisible Revolution, which surveyed more than 1,700 senior executives across manufacturing, energy, construction and utilities, found that 99 per cent of organisations anticipate significant reskilling needs as AI is adopted. The workforce transition and the technology transition are inseparable.
How AI Captures What Operators Know
AI systems deployed in industrial settings collect operational data continuously and at a level of granularity that no human observer could replicate. Over time, this record becomes a detailed account of how the plant has been run: the adjustments made by skilled operators, the sequences of decisions that preceded problems, the interventions that resolved them.
Machine learning models trained on this data begin to encode the patterns that experienced operators have internalised. They learn which subtle changes in process variables are meaningful and which are noise. They learn the relationships between inputs and outputs that a skilled operator knows intuitively but would struggle to articulate in a training manual.
ABB’s Global Digital Lead for Metals, Tarun Mathur, has described this directly, noting in analysis published by UNIDO in December 2025 that AI-powered systems could help preserve and transfer the operational expertise that retiring specialists risk taking with them. The system does not replicate human judgment fully, but it preserves a form of institutional memory that would otherwise be irretrievably lost.
The Performance Gap That Knowledge Closes
The value of addressing the tribal knowledge problem extends beyond continuity. Research on industrial performance consistently shows a wide distribution of outcomes across facilities that run nominally similar processes. Plants at the top of their performance range consume substantially less energy and produce fewer emissions per tonne of output than those operating at average levels.
Much of that gap is attributable to operational quality: the quality of decisions made by the people running the plant. Closing it, even partially, across the global industrial fleet would represent a material contribution to emissions reduction. AI systems that encode and apply the practices of high-performing operators have the potential to lift performance across facilities that would otherwise remain in the middle of the distribution.
The Limits of What Data Can Capture
Not all tacit knowledge transfers to data. Some of what experienced operators know relates to physical sensations, interpersonal dynamics or contextual judgements that do not manifest in sensor readings or control system logs. The goal of AI-assisted knowledge capture is not to reproduce human expertise in full, but to preserve what can be preserved and to reduce the gap between expert and non-expert performance.
This means that AI knowledge capture works best alongside, rather than instead of, structured human mentorship and knowledge transfer programmes. Companies that invest in documenting operational expertise before operators retire, and that create structured processes for transferring it to the next generation, are building a long-term asset. Those that treat workforce transitions as purely a human resources matter risk discovering the loss only after it has occurred.
A Strategic Priority for Industrial Operators
The industrial decarbonisation agenda depends on plants operating as efficiently as possible during the transition to lower-carbon technologies. Maintaining the quality of operational decision-making as experienced workforces change is one of the least visible challenges in that agenda. It is also one of the most consequential.
Organisations that treat operational knowledge as a strategic asset, and that deploy AI as part of a deliberate programme to capture and apply it, are building a form of resilience that will matter more as industrial transformation accelerates. The knowledge gap is a solvable problem. Solving it requires first recognising it as one.