February 14, 2025 Digital Twins: The Backbone of Effective AI Integration

In an industrial setting, integrating AI starts with a solid foundation – enter the digital twin.

By Jason Schern, Field CTO, Cognite

In the industrial environment, generative AI is becoming synonymous with transformation. It’s no longer a “nice to have” but is now critical to remaining competitive. However, for heavy-asset industries—such as oil and gas, manufacturing, and power and utilities—to take advantage of the transformative potential of generative AI, one critical obstacle must first be addressed: the industrial data problem.

A typical industrial facility routinely generates over 100,000 data points across more than 50 applications. These data points are not only high in volume but also incredibly diverse, encompassing both structured and unstructured formats. Structured data includes things like asset hierarchies and work orders, while unstructured data consists of items such as P&IDs (piping and instrumentation diagrams) and CAD drawings for the facility, often stored as PDF files. The data is also frequently siloed and spread across manual, disparate systems that don’t communicate effectively with one another. The crux of the data problem lies not with accessing data, but making sense of it and transforming it into actionable and usable insights.

Enter the digital twin.

Breaking down the digital twin

A digital twin is the aggregation of all possible data types and data sets, both historical and real-time, directly or indirectly related to a given physical asset or set of assets in an easily accessible, unified location. The collected data must be trusted and contextualized, linked in a way that mirrors the real world, and made consumable for a variety of use cases. In essence, digital twins bridge the gap between the digital and physical worlds and tell the story of what’s happening in real-time on the factory floor.

The ability to consolidate and contextualize data is particularly critical in industrial environments, where decisions must be made quickly and accurately. In these settings, you’re also dealing with heightened risks, such as complex machinery outputs, large-scale downtime, and safety hazards. Historically, the implementation of digital twins was limited to specific, isolated use cases. Today, with the rise of AI, they have become indispensable for industrial companies, as AI’s output is entirely reliant on context. Without a digital twin, AI is forced to make rapid decisions based on vague data points, lacking a clear understanding of the plant’s operations.

For large-scale industrial companies looking to integrate AI, the understanding and contextualization of data that’s unique to your operations – provided by a digital twin – serves as an essential foundation. The next step is moving beyond generalized language models to tailored solutions like industrial AI agents powered by digital twins. These agents use algorithms and data models optimized for domain-specific tasks, identifying patterns and anomalies unique to their field. The results are more accurate and actionable insights, scalable solutions for increasing data complexity, and improved productivity, safety, and operational efficiency.

How one global oil company integrated a digital twin

Let’s look at a specific example. In Japan, there’s been a decline in the working-age population due to a decreasing birth rate. By 2030, the working age population is expected to decrease by approximately 8%. Alongside this, the country has been working towards achieving carbon neutrality by 2050. Amid these trends, Japan-based Cosmo Oil anticipated that its refineries would need to be operated more autonomously, with smaller dedicated teams per site. In response, the company introduced a “digital twin of the refinery” which consolidated three separate refineries across the country into one virtual space.

Prior to the digital twin, Cosmo Oil’s data was scattered in various places, resulting in 70-80% of an engineer’s work dedicated to collecting the data across operations, maintenance, inspection records, and more. Additionally, each piece of data was managed within different formats – excel files, PDFs, text files, printed papers, etc.

The digital twin not only brought the data into one place, it utilized it to create a representation as if it were a twin of the virtual space. The use of the digital twin enabled collaborative maintenance. The company was able to obtain data from the three separate refineries and support maintenance operations for all three regardless of where they were. This resulted in increased efficiencies, improved working conditions, and maximized productivity across engineering teams. Had the company opted to go with a generalized large language model, the results would have been vague and indecisive – speaking to what could be happening out in the field instead of using hard data points unique to your operations to draw meaningful conclusions.

Digital twins and AI in action

As AI adoption in industrial environments continues to gain momentum, success increasingly depends on a strategic and thoughtful approach grounded in high-quality, relevant data. It begins with making sense of your data through a digital twin. By consolidating and contextualizing both structured and unstructured data, a digital twin creates a digital replica of what’s happening on the factory floor, updated in real time. The result is a strong foundation for your AI, which can now act on relevant, accurate, and contextualized data to produce actionable and meaningful decisions you can trust.

In industry, some of the biggest and most impactful opportunities for leveraging digital twin-powered AI are emerging in key areas. One example is collaborative maintenance, as seen with Cosmo Oil. Others include using a digital twin to integrate and contextualize disparate data and enhance predictive maintenance. By analyzing more data at an accelerated pace, digital twins help uncover hidden issues and address them before they escalate into larger problems.

We’ve also seen industrial companies use digital twins to simulate entire production lines, identify inefficiencies, and recommend optimizations. As more companies explore AI implementation, it’s essential to ground their strategies in data that’s unique to their operations. Many asset-heavy companies already possess a wealth of data – the key isn’t generating more but unlocking additional value from what’s existing. This is only made possible through a digital twin.

About the author:
Jason Schern serves as Cognite’s Field CTO. He has spent the last 25 years working with some of the largest manufacturing companies to dramatically improve their Data Operations and Machine Learning Analytics capabilities. Jason’s experiences led him to Cognite, passionate about the value and impact of trusted, accessible contextualized industrial data at scale.

Subscribe to Industry Today