For more than a decade, industrial companies have invested in predictive maintenance, using data to anticipate when equipment might fail. The next step, now being tested on the exhibition floor at Hannover Messe, is more ambitious: systems that not only forecast faults, but tell engineers what to do about them.
ABB is among those attempting to make that shift. At its stand with Microsoft, the company is presenting AI-enabled tools designed to move maintenance workflows beyond prediction towards what it describes as “prescriptive” intervention.
The distinction is subtle but significant. Predictive maintenance identifies the likelihood of failure; prescriptive maintenance adds a layer of interpretation, translating diagnostics into recommended actions. In theory, this reduces both downtime and the need for specialist expertise on site.
“Predictive maintenance tells you something may fail,” said David Lincoln, Digital Lead in ABB’s Measurement & Analytics division. “Prescriptive maintenance tells you what to do about it.”
ABB’s approach centres on integrating generative AI into its existing measurement and analytics systems. The division produces instrumentation used across industry to measure variables such as pressure, temperature, flow, and gas composition — often in regulated environments where uptime is critical.
At Hannover Messe, the company demonstrated how data from these devices can be combined with documentation and service records, then interpreted through Microsoft’s Copilot tools to produce natural language guidance for engineers.
The system, known as “My Measurement Assistant”, aggregates technical manuals, diagnostic data, and asset health reports into a single interface. Engineers can access it via QR codes on equipment, allowing them to retrieve device-specific information in real time.
Where earlier systems might have flagged an alarm and left the user to consult documentation, ABB’s model aims to close that gap. Diagnostic outputs are processed into health reports, which are then summarised by AI into actionable guidance.
The company says the approach can resolve around 80% of technical support queries without escalation, with a further 15% handled remotely and only a small minority requiring on-site intervention.
A recent update extends this model by linking condition monitoring systems directly into Microsoft Azure. Devices can now automatically generate alerts and send diagnostic reports into the cloud, where AI tools interpret the data and feed it into service workflows.
This begins to approximate a prescriptive system: faults are detected, analysed, and translated into recommended actions with minimal human input.
Yet ABB is cautious about how far this process can be automated. While more advanced “agentic” AI systems — capable of acting independently — are widely discussed, Lincoln said industrial deployment remains constrained by safety requirements and customer expectations.
“We are working towards more autonomous capabilities, but we are taking it step by step,” he said. “Customers need to understand and trust what the system is doing.”
