AI Is Changing Industrial Automation. It’s also changing the Role of the System Integrator.

A customer says they want to use AI in their operation. The first question probably shouldn’t be, “Which AI tool?” It should be: What does your industrial environment look like today?

Before AI can help predict a failure, identify process variability, optimize energy use, or recommend what an operator should do next, it needs to understand what is actually happening inside the operation. That means pulling information from PLCs, SCADA, historians, MES, maintenance systems, databases, and other sources. It also means knowing whether that data is accurate, whether equipment and processes are represented consistently, and whether information from different systems can actually be connected and understood together.

For many industrial organizations, that foundation is still a work in progress. Deloitte research found that nearly 70% of manufacturers identified data-related challenges, including data quality, contextualization, and validation, as significant obstacles to implementing AI. Its 2025 Smart Manufacturing Survey also found that only 29% of manufacturers surveyed were using AI or machine learning at the facility or network level, while another 23% were still piloting the technology.

A clear gap exists between exploring what AI could do and having an industrial environment ready to support it. For system integrators, that gap creates an opportunity to use their existing knowledge of industrial systems, data, equipment, and processes to help customers figure out what comes next.

AI Starts Long Before AI

When customers start talking about AI, the conversation can quickly jump to models, copilots, predictive analytics, or the newest capability they have seen demonstrated. On the plant floor, however, more fundamental questions come first. What problem are we trying to solve? What information would we need to solve it? Where does that information live today? How reliable is it? And does it have enough context to tell us what was actually happening when an event occurred?

Consider a manufacturer trying to use AI to understand why product quality changes from one run to another. The answer may depend on hundreds of variables across equipment, raw materials, process conditions, operator actions, environmental conditions, and previous production steps. Some of that information may already be available, but having the data doesn’t mean it is ready to be used together.

Before an AI model can find meaningful relationships, someone has to ensure the right information is collected, the systems can communicate, and the data accurately represents the process. System integrators already understand these challenges. The work may not have “AI” in the project name, but it can ultimately determine whether the AI initiative delivers anything useful.

Years of Data Does Not Automatically Mean AI-Ready

Most industrial organizations are not starting from zero. Many have years, or even decades, of operational information spread across historians, SCADA applications, PLCs, MES platforms, maintenance systems, spreadsheets, databases, and other applications installed throughout the life of a facility.

The challenge is that those systems were not necessarily designed to work together. One site may call an asset one thing while another uses a completely different naming convention. Two production lines performing nearly identical processes may have been engineered years apart and structured differently. Important process information may exist in the historian while the maintenance history needed to explain it sits somewhere else entirely.

People familiar with the operation may know how to navigate those differences, but AI does not automatically understand them. Deloitte’s 2025 Smart Manufacturing Survey found that 57% of manufacturers surveyed were already using data analytics at the facility or network level, compared with 29% using AI or machine learning at that same scale. That difference helps show why collecting and analyzing industrial data does not automatically make an operation ready for more advanced AI use cases.

For system integrators, this creates an opportunity to help customers make better use of the infrastructure and information they already have. Standardizing applications, cleaning up tag structures, connecting systems, improving data quality, and creating consistent asset models may not sound as exciting as deploying AI. Still, these are often what make industrial intelligence possible in the first place.

Context May Be One of the Most Valuable Things an SI Can Build

Imagine a pump begins drawing more current than usual. Knowing that the number changed is one thing, but understanding why it changed requires much more information. Was the pump running faster? Did flow change? Was a valve repositioned? Did inlet pressure change? Was maintenance recently performed? Is the increase unusual for this operating condition, or is it completely normal?

Those relationships are what turn individual data points into operational context, and that context becomes increasingly important when AI is expected to do more than detect an anomaly. If a customer wants AI to help explain what is happening or recommend what to do next, the system needs a much better understanding of the assets and processes behind the numbers.

This is where the way an industrial environment is engineered starts to matter even more. AVEVA System Platform, for example, allows equipment, processes, alarms, events, and historical information to be organized into structured operational models. Instead of treating thousands of tags as unrelated pieces of information, you can associate data with the equipment and processes it represents.

For system integrators, reusable object models have traditionally improved engineering consistency and made applications easier to deploy, maintain, and scale. As customers explore advanced analytics and AI, those same models can provide another important benefit by creating a more consistent layer of operational context that future technologies can use.

The Integration Challenge Is Getting Bigger

There was a time when integration projects could focus on getting the right information to the operator. Today, operational data has a much larger audience. Maintenance, engineering, continuous improvement, corporate operations, sustainability teams, and business systems may all need access to information that originated on the plant floor. Now analytics and AI applications are joining that list.

That changes the architecture conversation. The answer cannot simply be to send every piece of plant-floor data somewhere else. Industrial organizations need to determine what information should stay close to the process, what needs to be shared, how quickly it needs to move, how much context needs to travel with it, and who or what should be allowed to access it.

Platforms such as AVEVA CONNECT can help make trusted industrial information available beyond individual sites and applications, supporting enterprise visualization, analytics, collaboration, and AI use cases. DataOps technologies such as Crosser can help address another part of the challenge by connecting, processing, filtering, and contextualizing information across edge, on-premises, and cloud environments.

For a system integrator, the value is not simply knowing how to connect these technologies. It is understanding how the architecture should work for the operation. A high-frequency process variable may need to remain at the edge, while an abnormal event may need to be processed locally and immediately shared with central operations. A calculated production KPI may need to be available across multiple sites. Those decisions depend on the customer’s process, infrastructure, security requirements, and business goals, which is why no single AI-ready architecture works for everyone.

AI Makes Industrial Knowledge More Important, Not Less

Plenty of discussion focuses on what AI could eventually automate, but industrial operations are not generic environments. A recommendation that makes sense for one production line may not make sense for another, and the same piece of equipment can behave differently depending on the product, process, operating conditions, or surrounding assets.

System integrators bring an understanding of the control system, equipment, network, software, and often the customer’s process itself. They know why a five-second delay may be irrelevant in one application and unacceptable in another. They understand why simply connecting a system to the cloud is not an architecture strategy, and why changing something upstream can have consequences elsewhere in the operation.

That experience becomes even more valuable when AI begins influencing operational decisions. Deloitte found that 78% of manufacturers surveyed were allocating more than 20% of their overall improvement budgets toward smart manufacturing initiatives, with data analytics and AI among their technology investment priorities. As more of that investment moves from pilots and experimentation toward real operational use cases, customers will need people who can connect those ideas back to how their facilities actually operate.

The SI Opportunity Is Bigger Than an AI Project

For system integrators, the opportunity is not necessarily to become an AI company. It is to become the partner that helps customers determine whether their operations are ready to support AI and what needs to change if they are not.

That conversation can begin by looking at the outcomes the customer is trying to improve and working backward. What systems are already in place? Where does the necessary data live? Can that data be trusted? Is enough historical information available? Does it have the right context? Can information move securely between the systems that need it? Answering those questions creates a much more practical starting point than choosing an AI technology first and trying to make the operation fit around it.

The answers may lead to a historian modernization project, a SCADA upgrade, better asset models, improved system connectivity, a DataOps strategy, stronger data governance, or a new edge-to-enterprise architecture. In other words, the path to AI may begin with many of the things system integrators already do. What changes is the conversation around that work and the outcomes customers expect.

Instead of only helping a customer deploy the next system, SIs can help customers understand how the systems, infrastructure, and industrial data they already have can support where they want to go next. AI may start that conversation, but building the foundation that makes it useful is where system integrators can have an even bigger impact.

Help Your Customers Get Ready for What’s Next

AI is creating new conversations with industrial customers, and system integrators have an opportunity to lead them. InSource works alongside SIs with the technology, expertise, training, and support needed to strengthen industrial architectures, solve data challenges, and help customers prepare for what comes next.

Partner With InSource