Industrial AI has a context problem – and it’s time to solve it

For much of the past decade, industrial companies have been dutifully collecting data – almost as if it were a badge of honour. More data leads to better decisions – or so the thinking went.

The scale is staggering: nearly 40 billion connected IoT devices are expected by 2030, alongside a global industrial IoT market projected to exceed $1.6 trillion.

Across energy, mining, utilities and manufacturing, organizations invested heavily in connectivity, cloud, data lakes, and digitization programmes designed to aggregate operational information at rapid scale.

Yet many of those initiatives ultimately disappointed because the operational context that gave the data meaning was not preserved. When aggregating information, companies focused on sheer volume rather than quality.

In industrial operations, context is everything. Timestamps, engineering relationships, asset hierarchies, units of measure, and process lineage are the mechanisms through which industrial systems become understandable, trustworthy and actionable.

A pressure reading without equipment lineage, maintenance history or process context may technically be “data”, but its operational value is diminished. As a result, many ambitious industrial data lake strategies struggled to deliver worthwhile business outcomes.

In some cases, data lakes evolved into what many industrial operators now call “data swamps” – vast murky pools of information that lack the coherence needed to support reliable decision-making.

This challenge is becoming ever more significant as industrial companies accelerate their adoption of AI. For all the furore surrounding generative AI and advanced analytics, many organizations are beginning to recognize that industrial AI does not fundamentally have a data quantity problem. It has a context and understanding problem.

The next phase of industrial AI will therefore depend less on data volume and more on whether organizations can preserve the digital thread between assets, processes, people and decisions.

This shift requires a move towards what technology leaders call “industrial intelligence” – the convergence of data-driven insights, human expertise and AI into a unified operational decision-making framework that helps organizations move from insight to action more effectively. Importantly, this emerging model is not about replacing human expertise – it is about augmenting it.

Industrial operations are highly complex environments where tacit operational knowledge, engineering judgement and real-world process understanding continue to matter enormously. Experienced operators often possess decades of accumulated knowledge that cannot easily be replicated through isolated AI models alone.

At the same time, many industrial sectors face demographic and workforce pressures. Swathes of experienced workers are approaching retirement, while companies increasingly seek to accelerate the productivity of newer employees operating in highly sophisticated technical environments.

The challenge is therefore twofold: how to preserve institutional operational knowledge, while simultaneously enabling faster, better decision-making across increasingly volatile industrial systems.

This is where the evolution of the digital twin is becoming particularly important. For years, digital twins were seen primarily as visual representations of physical assets or facilities. But today that concept is transforming into something much broader: an interconnected intelligence system capable of integrating engineering, operational and enterprise-level information into a contextualized environment that can be understood and acted upon by both people and AI.

Rather than functioning as a static model, the digital twin is evolving into a dynamic operational layer that can unify information such as time-series operational data, process data, engineering models, asset relationships, maintenance records, and workflows. Increasingly, these environments also incorporate documents, procedures and other unstructured information that provide critical operational context.

The convergence of these elements creates a far more complete representation of operations and allows organizations to move closer to near real-time operational visibility.

Moreover, the modern digital twin preserves relationships between systems, ensuring that industrial information retains contextual integrity as it moves across facilities, business units and decision-making layers.

Often, industrial firms are working with inconsistent naming conventions and fragmented operational structures. For example, one site may label equipment differently, while data structures may vary across geographies.

This is where knowledge graphs and unified namespaces underpinning the digital twin become important – creating relational structures that allow industrial information to be connected coherently across assets, facilities and operational domains, helping AI systems understand relationships rather than isolated data points.

As industrial systems grow more interconnected and operational decisions become more dependent on rapid information flows, this capability is becoming increasingly important.

For example, SCG Chemicals, one of the largest petrochemical companies in Asia, pursued a digital transformation initiative to harness data and build an advanced asset performance management solution to monitor critical assets and predict failure.

By leveraging trusted, contextualized real-time data blended with analytics, teams are now able to predict equipment health and optimize performance more readily. This approach helped increase plant reliability from 98% to nearly 100%, delivering a ninefold return on investment within six months.

In a world where industrial operators have faced repeated disruptions stemming from geopolitical tensions, supply-chain instability and energy market volatility, responsiveness and near real-time visibility across the entire organizational ecosystem becomes crucial.

This is where advanced contextualized digital twins are beginning to function as resilience infrastructure. By preserving the digital thread between assets, processes and decision-makers, industrial intelligence systems can help organizations identify disruptions earlier, model operational impacts more accurately and coordinate responses more effectively. As industrial AI matures, this shared operational context will become increasingly important for coordinating decisions and actions across functions, sites and workflows.

The industrial sector is therefore entering a new phase of digital transformation. The first phase focused heavily on connectivity and data aggregation. The next and most consequential era is industrial AI. Crucially, this isn’t about bolting AI onto broken data foundations. It is AI built on trusted, contextualized operational intelligence.