What is the problem with production data in industry?
In Dutch industry, machines run and produce data. Every shift, every batch, every standstill leaves a trail. But that trail dead-ends in a PLC, a historian, or a local Excel file on the production planner's laptop. The ERP system receives a manual entry at the end of the day. The planner works on assumptions. The quality manager exports a report and pastes it into a presentation. Meanwhile, the actual production data sits somewhere else entirely. According to UWV research (Industrie in beeld, February 2026), around 80% of industrial companies are experiencing greater economic uncertainty, while labour shortages remain one of the biggest challenges. That makes this kind of duplicate work especially costly: the people retyping data are the same people needed for the real work.
Why does a standard MES not solve this automatically?
A standard MES package does not resolve data disconnection on its own. Most packages are built for a generic factory, not for your specific combination of machines, raw materials, customer specifications, and planning logic. Implementations take a long time, configurations become compromises, and after two years the system is already half-outdated relative to how the operation has grown. The real problem is that the MES still does not connect to the ERP, the quality registration system, or the purchasing department. So integrations are built, scripts are run, and Excel files are maintained as a safety net. ING expects Dutch industry to grow by around 2% in 2026, driven primarily by high-value manufacturing sectors. Those are exactly the companies where production data is most complex and can yield the most valuable insight. And they are precisely where a generic package delivers the least return.
Which processes break down most often?
In practice, the same bottlenecks appear repeatedly. Work order processing: the order comes from the ERP, gets printed, filled in on the shop floor, and then manually retyped. Quality registration: measurement values are recorded on paper or in a local spreadsheet, and only entered into the quality system later. Production reporting: shift handover notes are passed on verbally or via an app that is not connected to the planning system. Each of these processes relies on people acting as a human link between systems that should be handling this themselves. That costs time, introduces errors, and makes real-time course correction impossible. A production manager who only sees the next morning that a batch fell outside tolerances the previous evening cannot use that information for the decision that needed to be made the night before.
When is an AI layer the right solution -- and when is it not?
An AI layer on top of existing systems helps when core processes are already captured digitally but information is not flowing through. AI Workers can then read documents, forward work orders, flag deviations, and generate notifications without a person touching every step. This works well for companies that do not want to replace their ERP or MES but do want to eliminate manual data entry and manual linking. It does not work when the underlying data is structurally absent or when the core system itself is so outdated that there is no reliable foundation to build on. In that case, rebuilding the core system is the more honest choice. A system built AI-native, with production logic at its centre, gives the operation a foundation that scales with it. The choice between these two routes depends on the state of the current systems, the complexity of the operation, and whether the existing architecture is scalable at all.
What does it deliver when production data actually flows?
When work orders are forwarded automatically, quality data is available in real time, and the planner works from current capacity data, the operation changes in a tangible way. Not because some AI miracle has been deployed, but because information arrives at the right place at the right time. Adjustments can be made within the shift, not the following day. Deviations are caught before they render a batch unusable. The production planner spends time on actual planning instead of synchronising systems. In a sector where labour shortages are a structural reality, that is not a luxury. It is simply the most efficient way to get the best out of the people you have.