MES is back at the center of manufacturing strategy
The more manufacturers invest in AI, digital twins and advanced analytics, the more obvious a basic weakness becomes. These technologies can generate insights at extraordinary speed, but they cannot reliably influence production unless they understand what is being made, which process is running and what conditions surround the data. That is forcing manufacturing execution systems (MES) back into the strategic conversation.
Leonor Marques, Architecture & Advocacy Director at Critical Manufacturing, sees the resurgence of MES as a response to the pace and complexity of modern production. Smaller batches, more frequent changeovers, higher customization and deeper automation have made the old model of reviewing performance at the end of a shift increasingly inadequate. MES is moving beyond production enforcement because it holds the operational context that links plant data with decisions.
“A lot of the important decisions used to be taken elsewhere, whether in meetings, ERP systems or planning tools,” Marques says. “Now production changes faster, automation is creating far more data and manufacturers need to understand what that data means in the context of what is being produced. MES is therefore becoming more of an operational platform than a simple execution system.”
The value lies between insight and action
Manufacturers have become better at seeing what is happening, but visibility alone does not create value. Dashboards are richer, reports arrive faster and AI can surface patterns in seconds, yet the final step from insight to intervention often remains weak. The problem is not always a lack of information, but the absence of a trusted route from analysis into the production workflow.
MES occupies that position because it sits between IT and OT. It knows which order is running, what material is being consumed, which recipe is active, how equipment is configured and where a unit sits in the process. A sensor reading becomes more useful when it can be connected to the product, machine state and operating conditions that produced it.
“The point is not simply to generate insights, but to act on them,” Marques adds. “If the insight is wrong, it can be worse than having no insight because people will make decisions based on incorrect information. A modern MES provides the runtime context needed to decide faster and with greater confidence.”
That distinction is particularly important for industrial AI. A model can produce a confident answer from fragmented or poorly contextualized data, and the risk increases when the recommendation is fed directly into a fast-moving process. MES can provide the production history, relationships and event context needed to test whether the answer reflects what is really happening.
The same principle applies to digital twins. Without continuous synchronization with production, a virtual model may still be valuable for simulation, but it cannot represent the current state of the factory. MES provides the runtime information that allows a twin to support operational decisions rather than remain an isolated engineering tool.
Modern MES also addresses the fragmentation created by years of buying specialist applications for individual problems. One production event may reduce OEE, increase scrap and trigger a quality deviation, yet separate systems can treat these as unrelated issues. “If OEE, scrap and quality are analyzed in separate applications, the same event can appear as several different problems,” Marques explains. “You may end up looking for different causes and trying to solve each symptom independently. The value of MES is that it can connect those shop-floor events and help the operation understand the situation as a whole.”
Strategy is now an architecture decision
The strategic question is no longer whether an MES can satisfy a fixed list of requirements. It is whether the platform can evolve as products, processes, technologies and business priorities change. A system selected only against today’s checklist may become another constraint when the manufacturer adds new analytics, plants or AI capabilities.
“If manufacturers select MES only against the features they need today, they are missing the strategic issue,” Marques notes. “Those requirements will change in two or three years, so they need to look at the architecture and its ability to evolve. Otherwise, they will keep patching individual problems instead of building a foundation for the operation as a whole.”
This becomes especially important for multi-site manufacturers, where apparently comparable plants may be operating with different systems, definitions and data structures. The problem is not simply inconsistency between sites, but the risk of making enterprise decisions from information that cannot be compared on equal terms.
“You may have several factories producing the same product, but if each site uses different systems and its own data architecture, the comparison will never be completely reliable,” Marques says. “The danger is that decisions are then made for the wider company using information that is still shaped by the way one individual plant works.”
A strategic MES program can create a common operational language across the network. It does not require every factory to operate identically, but it does require shared definitions for products, assets, events, quality states and performance measures. Once those foundations are in place, manufacturers can compare plants with greater confidence and transfer successful practices more effectively.
The same architecture can help preserve operational knowledge. If one factory encounters a pattern already seen elsewhere, the system may identify the similarity, retrieve the previous response and show which intervention worked. “Imagine the same product is being made in several factories and a problem appears at one site that has already occurred somewhere else,” Marques continues. “The system could recognize the pattern, show what was done previously and indicate what worked and what did not. That would help the new site make a better and faster decision.”
Cloud-native and modular architectures can accelerate this development, but composability cannot mean assembling an unrestricted collection of applications. Modular services help manufacturers adapt faster only when the data structures, governance and event models remain consistent. “You need to make sure there is a connection between the data structures so everything can communicate in real time,” Marques says. “Composable elements are useful, but there still has to be a common layer of governance across the data and AI architecture. The objective is not a set of separate applications, but a framework that works together.”
Giving operators decisions rather than dashboards
The workforce argument is central to the renewed importance of MES. Experienced operators understand products and processes, but they should not need SQL, reporting or data engineering expertise to interrogate production information. Natural-language interfaces and embedded analytics can allow people on the shop floor to ask questions directly and receive answers grounded in the production model.
“If the insights are embedded in the system, supervisors and operators can ask questions in natural language rather than waiting for an analyst or SQL expert,” Marques adds. “That allows people who know the process, but not necessarily the technology, to understand what happened and act more quickly.”
The next stage will move MES from decision support toward selective autonomy. Agentic AI may allow systems to recognize familiar production conditions, evaluate available responses and act within defined limits without waiting for human approval for every routine intervention. This will be a controlled progression rather than an abrupt move toward fully autonomous factories.
“For the most frequent situations and patterns, MES will start acting by itself,” Marques explains. “Human decisions will remain on the table, especially in areas such as medical device manufacturing where the consequences are critical. What will change is the way people intervene and the point at which their judgment is required.”
That model keeps accountability with people while reducing the burden of routine interpretation. Operators may spend less time responding to repetitive alerts and more time setting boundaries, validating exceptions and improving the logic used by the system.
MES can also make sustainability part of production rather than a parallel reporting exercise. Statistical process control can detect machine deterioration before it creates scrap, while time constraints can ensure prepared materials are used before they expire. Better coordination can reduce unnecessary equipment runtime, rework and energy consumption.
“If the system recognizes a pattern of deterioration in a machine, it can stop the equipment before it starts producing bad material,” Marques concludes. “That saves energy, materials and components that would otherwise be assembled into products that cannot be used. MES can also control preparation windows so that materials are not wasted.”
The renewed strategic value of MES comes from its ability to connect ambitions that manufacturers often pursue separately. AI needs production context, digital twins need synchronization, multi-site operations need common definitions and workers need information they can act upon. MES is becoming the operational layer through which those capabilities can influence what happens next, rather than simply explain what happened before.
Manufacturers that continue to treat MES as a narrow system for dispatching work and recording completion may keep production moving. Those that treat it as an operational data and orchestration platform will be better placed to connect plants, preserve knowledge, support workers and scale AI. The strategic test is whether the architecture can explain what is happening, preserve the context needed to trust the answer and help the operations act before the value of that answer disappears.
