The Worker That Never Clocks Out: How Digital Workers Are Reshaping Industrial Maintenance

Somewhere in a manufacturing facility right now, a maintenance team lead is doing a job that does not appear on any org chart. Before the first machine is switched on, before the morning briefing, before the first coffee, that person is building a mental picture of the night: what broke, what was fixed, what the shift logs say, what the notification queue looks like. And, and which urgent action takes top priority when the rest of the team starts their day. At Berkvens Doorsystems, a Dutch door manufacturer operating across three production facilities, that process consumed up to an hour every single morning. Per team lead. Every day. 

It is not a rounding error. It is not a process inefficiency waiting to be optimized. It is what maintenance leadership looks like in industrial organizations, and it has looked this way for decades. What is changing now is that the work is no longer entirely human. 

A Workforce Challenge That Data Alone Cannot Solve 

The industrial maintenance sector is facing a structural workforce problem that has been building for years. Ultimo’s Maintenance Trend Report found that 63 percent of industrial organizations are struggling with an aging workforce, a figure that captures something more troubling than a headcount shortfall. When experienced technicians retire, they take with them something that does not live in any EAM platform – the diagnostic intuition that comes from years of watching specific machines behave in specific ways. That knowledge is not documented. It is not transferable on a spreadsheet. In many organizations, it simply walks out the door. 

At the same time, the volume of data generated by modern asset operations has grown well beyond what any team can process manually. Work orders, shift logs, maintenance histories, equipment notifications, Internet of Things (IoT) and sensor data, and environmental, health, and safety (EHS) records pile up across systems, generating a picture of operational health that is theoretically complete and practically unreadable. The gap between having data and acting on it is where operational value gets lost. It is also where a new category of artificial intelligence (AI) tool is beginning to close the distance. 

What a Digital Worker Actually Does 

The term “digital worker” is doing a lot of work in vendor conversations right now, and it is worth being precise about what it means in an industrial maintenance context. A digital worker is not a chatbot that answers questions about equipment specs, and it is not a dashboard that surfaces key performance indicators (KPIs). It is an AI system embedded in operational workflows that can independently monitor data, initiate actions, and complete multi-step tasks without requiring a human to structure each one. 

In practice, for a maintenance planner, that might mean a system that pulls together overnight notifications, cross-references them against open work orders and maintenance history, and delivers a structured briefing by the time the morning shift begins. For a technician, it could mean a tool that surfaces similar past failures when a new fault code appears, narrowing the diagnostic search before a wrench is picked up. For a safety manager, it means incident reporting workflows that run consistently because the process is embedded in the system, not dependent on someone remembering to follow procedure under pressure. 

The common thread is that digital workers handle the high-volume operational tasks that require attention and context but not continuous human analysis, and that have historically consumed the time of the people best qualified to do something more valuable with it. 

From Theory to Shop Floor: The Berkvens Experience 

Berkvens Doorsystems, a family-run manufacturer headquartered in Someren, Netherlands, with nearly a century of tradition behind it, offers one of the clearest early accounts of what this shift looks like in practice. The company had already made a foundational investment in an enterprise asset management platform to bring structure to maintenance planning, work order management, and equipment tracking. That system provided visibility. What it could not do was synthesize the data fast enough to be useful at the start of a shift. 

“With AI, all relevant information is automatically summarized and combined, saving each team lead 30 to 60 minutes daily during the start of the day and matching our own analysis by more than 95 percent,” says Stefan van Bussel, Teamlead Technical Services at Berkvens Doorsystems. That accuracy figure matters as much as the timesaving. A summary that requires extensive correction is not actually a time- saving. A summary that matches human analysis at better than 95 percent is a genuine transfer of cognitive work. 

Translated across a team of technical leads, that recovery represents between 125 and 250 hours per person per year, representing an estimated annual value of €12,500 to €25,000 per team lead, depending on labor costs. These are not projected returns from a business case presentation. They are reported outcomes from a deployment already in operation. 

Efficiency Is Not the Main Story 

The productivity gains are measurable and significant. But Berkvens Doorsystems’ experience suggests they are not the most strategically important part of what digital workers deliver.  

Jeroen Wijnen, Maintenance and Installations Leader at Berkvens Doorsystems, puts it directly: “By intelligently combining data from different sources, new insights emerge, enabling teams to set better priorities, identify structural issues, and carry out more targeted maintenance and improvements.” 

What Wijnen is describing is a qualitative shift in what gets noticed. When maintenance data streams such as work orders, logbooks, equipment notifications, and maintenance histories are reviewed in isolation by different people at different times, structural patterns are genuinely hard to detect. A fault that recurs every six weeks across a specific equipment type looks like bad luck when seen one incident at a time. Seen across a connected data set, it looks like a systemic issue with a root cause. Those two framings lead to completely different responses. 

For Berkvens Doorsystems, this revealed that certain equipment issues were recurring structurally rather than randomly, a finding that changed how the maintenance team approached planning and where it directed improvement efforts. The AI did not make that decision. It made the pattern visible in a way that no manual review process reliably could. 

Human Control Is Not Optional 

One of the more persistent anxieties about agentic AI in operational settings is the question of oversight: what happens when an AI system takes an action that a human would not have sanctioned? It is a legitimate concern, and in industrial maintenance, where decisions carry safety and compliance implications, it deserves a direct answer. 

The design principle that appears to be emerging in well-built industrial AI deployments is a distinction between autonomous action on non-disruptive tasks and human-in-control requirements for higher-risk decisions. Routine reporting, data synthesis, and pattern identification can run without interrupting a technician’s workflow. Decisions that affect maintenance schedules, equipment availability, or safety compliance trigger human review before any action is taken. 

This architecture matters because it determines whether AI in the workplace is genuinely augmenting human judgment or quietly replacing it in areas where oversight would be appropriate. The former makes organizations more capable. The latter creates liability. The distinction is not semantic; it is a design requirement that industrial organizations should examine carefully when evaluating any AI deployment in their operations. 

The Data Foundation Question 

Berkvens Doorsystems offers one of the more honest assessments of what makes an AI deployment work, and it has nothing to do with the AI itself. The company’s consistent message is that years of disciplined, structured use of their EAM platform created the data foundation that made digital workers viable. “Poor data hygiene does not get corrected by AI; it gets amplified,” is how the team frames it. 

This is the practical readiness test that industrial organizations should apply before investing in AI for maintenance operations. The technology is capable. Whether a specific organization can benefit from it depends on whether the underlying data is structured, consistent, and accurate enough to produce reliable outputs. Organizations that have treated asset management as a compliance exercise rather than an operational discipline will find that AI surfaces the quality of their data, not a corrected version of it. 

For organizations that have made that foundational investment, the barrier to productive AI deployment is lower than most assume. The data already exists. The workflows already exist. The value is in connecting them with intelligence that operates at a speed and scale no team could achieve manually. 

What Comes After the Morning Briefing 

The near-term direction for digital workers in industrial maintenance is moving toward troubleshooting support: systems that not only identify what has failed but draw on the accumulated failure history of similar equipment to surface probable causes and suggested diagnostic steps. For organizations operating across multiple facilities, this carries particular significance. 

A technician facing an unfamiliar fault on a machine at a site where the relevant expert is unavailable currently relies on whatever documentation exists and whichever colleagues can be reached. A digital worker with access to that equipment’s full failure history, cross-referenced against similar incidents across all of the organization’s sites can narrow the diagnostic field before anyone picks up a tool. That is not replacing experience. It is making accumulated organizational experience accessible to any technician, at any site, in the moment it is needed. 

This is the specific gap that the industrial sector’s workforce challenge creates. The retirement of experienced technicians does not just remove headcount; it removes pattern recognition that took decades to develop. Digital workers, trained on operational data, offer a mechanism for preserving some of that institutional knowledge in a form that remains useful after its original owners have left. 

The New Operational Baseline 

The maintenance engineer arriving for a morning shift has always relied on experience, instinct, and whatever handover notes the night team remembered to write. That has been the operational baseline for as long as industrial maintenance has existed. It is changing, not because AI is replacing experienced professionals, but because the volume and complexity of operational data has grown past the point where human review alone can extract its full value. 

The organizations beginning to move on this are not doing so because they have resolved every philosophical question about AI in the workplace. They are moving because the gap between what their data contains and what their teams can process is becoming a competitive constraint. Digital workers are not the answer to that problem in the sense of solving it permanently. They are a mechanism for closing the gap continuously, at scale, without adding headcount. 

What industrial organizations considering this shift need to evaluate is not whether the technology works. Early evidence suggests it does. The questions that matter are about data readiness, about which workflows carry the highest value for augmentation, about how oversight and human control are designed into the system from the start, and about what training their teams will need to work effectively alongside a new kind of colleague. 

Those are not easy questions. But they are the right ones to be asking and asking them now puts organizations ahead of the longer and more difficult conversation that comes when falling behind.