Manufacturers turn to AI as maintenance know-how fades

IoT Analytics has published research arguing that manufacturers are losing maintenance knowledge as experienced technicians retire, with AI tools emerging as one response.

The research links the issue to the broader cost of unplanned industrial downtime, which it estimates at USD $1 trillion a year worldwide. It argues that the maintenance discussion is shifting from predicting failures to preserving and applying know-how that has often remained undocumented on factory floors.

That shift comes as maintenance teams lose long-serving staff who hold practical repair knowledge built up over years of work on industrial assets. The analysis says technology suppliers are trying to address the gap with software and systems designed to capture, structure, and use maintenance expertise in day-to-day workflows.

The findings highlight five themes shaping the market. Data quality is identified as the main obstacle to wider use of AI in maintenance, while wireless sensing is widening the pool of maintenance information available to companies.

The report also finds that asset-health ecosystems are starting to take shape as suppliers connect tools and data sources more closely. At the same time, vendors are moving beyond predictive maintenance by developing platforms that suggest what action should follow once a potential fault is identified.

Cloud use remains a sensitive issue for many industrial operators. Concerns about cybersecurity and data sovereignty continue to slow adoption, prompting vendors to offer different system designs to address those objections.

Knowledge gap

The central argument is that the maintenance challenge is no longer only technical. It is also organisational, because companies risk losing repair knowledge that was never written down, standardised, or shared widely enough to survive workforce turnover.

That matters because the loss of expertise can lengthen mean time to repair, especially when less experienced workers must diagnose faults without access to the judgment of veteran technicians. The analysis suggests that if companies fail to preserve that knowledge, the effects could go beyond staffing pressure and feed directly into downtime and cost.

Knud Lasse Lueth, Chief Executive Officer at IoT Analytics, set out that view in the report's commentary. "For the last decade, many of my discussions around smart maintenance have focused on how manufacturers can predict machine failures before they happen. That remains important, but manufacturers now face a more structural challenge: experienced technicians are retiring, and much of their practical maintenance knowledge has never been formally documented. We believe AI can help close this gap by capturing expert know-how and making it available to frontline teams when they need it. But there is a real sense of urgency. These systems take time to build, validate, and embed into daily workflows, while some of the most experienced people are already leaving the workforce," Lueth said.

Data limits

The report places particular weight on the condition of industrial data. It argues that AI systems in maintenance depend on reliable information from machines, sensors, historical service records, and operator inputs, and that weak or inconsistent data can undermine trust in the results.

For that reason, wider deployment may depend less on enthusiasm for AI itself than on the quality of maintenance records manufacturers already hold. In practice, companies that digitised unevenly or kept fragmented service histories may find it harder to turn AI into a useful tool for frontline maintenance teams.

Wireless sensing is presented as one way to broaden the available evidence base. By collecting more operational information from equipment that was previously harder to monitor continuously, such systems could help companies build a larger store of machine and maintenance knowledge over time.

The study also points to a gradual broadening of maintenance software beyond fault detection. Instead of stopping at warning users that an asset may fail, newer tools are being developed to recommend the next step, linking insight to maintenance action.

Industrial caution

Even so, industrial buyers are not moving without reservations. Cloud hesitancy remains a factor in the market, particularly where operators worry about exposing sensitive operational data or storing it in ways that raise regulatory or sovereignty concerns.

That caution helps explain why suppliers are offering varied deployment approaches rather than pushing a single model. The aim is to make AI and maintenance software acceptable to manufacturers that want digital tools but remain reluctant to move all relevant data into the cloud.

The report frames the issue as a race against time for industrial companies with ageing technical workforces. "For the last decade, many of my discussions around smart maintenance have focused on how manufacturers can predict machine failures before they happen. That remains important, but manufacturers now face a more structural challenge: experienced technicians are retiring, and much of their practical maintenance knowledge has never been formally documented. We believe AI can help close this gap by capturing expert know-how and making it available to frontline teams when they need it. But there is a real sense of urgency. These systems take time to build, validate, and embed into daily workflows, while some of the most experienced people are already leaving the workforce," Lueth said.