For engineering assistant Mangleswaran Mahalingam, getting a stalled machine up and running used to mean leafing through hundreds of pages in a manual to find an error code and its recommended fix. If that didn’t work, he’d have to track down a colleague who’d resolved a similar snag before. It was a painstaking process, often involving trial and error.
Today, however, he can access that collective knowledge in seconds.
Since October 2025, technicians overseeing hundreds of machines at Rockwell Automation’s Singapore site have been using the GenAI-Powered Maintenance Copilot, an in-house AI assistant trained on its workers’ expert knowledge, data from its manufacturing software and the machines’ instruction manuals.
Powered by Microsoft Azure and built using Azure OpenAI in Microsoft Foundry, this AI tool runs on Open AI’s large language model GPT-5.4-mini, helping the maintenance crew diagnose and troubleshoot problems faster.
“I used to have to look for colleagues or senior technicians and ask them if they’d seen that error before,” says Mahalingam. “Now, I can type my questions in the copilot on my tablet and get answers from it. It’s quicker and easier.”
Headquartered in Wisconsin, industrial automation giant Rockwell supplies hardware and technology that help companies across diverse industries coordinate their complex operations, be it electronic components to automate how luggage is routed in an airport, or a control system to blend and bottle a soft drink.
Quick access to expert insights
When a machine error is detected at the Singapore factory, the AI assistant searches for expert insights on it from a database created by veteran Rockwell engineers who each have 20 to 30 years’ experience on the shop floor. It organizes the information by symptom, cause and reaction, so the technician can immediately see what the top likely causes of the error are, be it a worn-out gear or a faulty sensor.
The technician can also read what other colleagues have said about this issue, and whether it’s occurred in other production lines. Meanwhile, they can still access information from those thick instruction manuals, but by posing natural language queries to the copilot, which responds with a user-friendly, step-by-step guide.
“The key thing we want to solve is the loss of knowledge, what we call the tribal knowledge,” says Singapore plant director Li Wang, referring to decades of tacit expertise that leave Rockwell when workers retire. “We also want consistency in troubleshooting, so there’s a standard approach to a problem regardless of who is working on it.”
This consistent, targeted approach has lowered their machines’ downtime by 33 percent. They spend less on servicing and spare parts, with Rockwell’s internal tracking showing costs are down by about 25 percent
Having all that knowledge residing in the copilot also helps workers learn the ropes faster.
Mahalingam recalls feeling worried when he first stepped foot in the factory in 2021, because, “I didn’t know anything about the machines.” It took him more than a year before he felt confident enough to handle night shifts, when fewer colleagues are on duty. Now he oversees about 50 machines and has learned to access the AI using smart augmented reality goggles.
Rockwell estimates that onboarding time for new workers has been shortened with the AI assistant.
“Instead of it taking nine months for them to learn how to troubleshoot hundreds of machines, it takes three months. Which is a huge improvement,” says Bob Buttermore, Rockwell’s chief supply chain officer. “And AI was the catalyst for that.”
The Singapore facility makes industrial automation equipment like controllers and networking components for high-growth sectors such as pharmaceuticals, automotive and semiconductors.
In June, the plant was designated as a World Economic Forum global lighthouse site. In its citation, the WEF noted that the site “increased units per person-hour by 43%, reduced defects by 35% and shortened time-to-competency by 67%.”
The maintenance copilot was one of several AI-driven capabilities implemented at the facility that contributed to these efficiencies, says plant director Wang. Others include a multi-agent quality assurance system that detects and addresses product defects on the assembly line in real time, and a predictive maintenance system that helps prevent unplanned downtime of critical equipment.
“These AI solutions demonstrate how we can combine our operational and automation expertise with AI technology to produce innovation,” says Patrick Dey, vice president in charge of data, analytics and AI innovation at Rockwell. “This innovation doesn’t just improve our own operations; we in turn use it to help our customers accelerate their digital transformation efforts.”
Dey says his team built these AI tools leveraging the breadth of the Microsoft technology stack. This included Azure OpenAI Service and Azure AI Search for conversational experiences and intelligent reasoning, and Azure’s security, identity and governance services to ensure enterprise-grade protection and compliance.
Wang sees these solutions, resulting from close collaboration between Rockwell’s operations and tech teams, as just the start of their AI journey.
“They are becoming more of a co-worker to our team. To us, these applications are more than just a solution. They are also leading the whole team into a new world.”
That future Rockwell chief information officer, is one where AI is in the loop for every piece of work. “This productivity is going to allow people to do more and create new possibilities,” he says. “For AI to be successful, we need good data, a trained workforce and a culture focused on learning and outcomes.”
The Singapore factory also acts as a model manufacturing site for Rockwell’s global operations and clientele. Lessons learned here can be used to drive digital transformation for customers and in other parts of Rockwell, which has over 25 plants worldwide and more than 26,000 employees. The maintenance copilot, for instance, will be rolled out in Rockwell’s manufacturing plant in Twinsburg, Ohio, as well as in Poland and Mexico, says Buttermore.
Key to this is its technology architecture that can be scaled globally.
“Microsoft’s cloud-native architecture supports expansion from a single factory to global deployments,” says Dey. “Integration within the Microsoft ecosystem gives us seamless connectivity across data, AI, applications and collaboration tools.”
For Rockwell, using AI to drive efficiency and autonomy positions it as a factory of the future, and demonstrates what it can do for customers.
“Ultimately, we want to be a showcase for our customers,” says Buttermore. “By deploying, testing and scaling these solutions in our own facilities first, we’re able to bring proven approaches to manufacturers looking to improve productivity, resilience and workforce effectiveness.”
Top photo: Machine technicians at Rockwell Automation’s Singapore site use an AI-powered maintenance copilot to diagnose and troubleshoot problems. (Photos courtesy of Rockwell Automation)
Lim Ai Leen reports on AI for Microsoftformerly associate foreign editor at The Straits Times in Singapore and still pens an occasional weekend column. Contact her on LinkedIn
