AI Assistants Join the Factory Floor

The basic machine For grinding a steel ball bearing, the same has been the same for around 1900, but the manufacturers have everything around the surroundings. Today the process is powered by a conveyor belt and mostly automatically. The most urgent task for humans is to find out when things go wrong – and even that could be handed over soon AI.

The Schaeffler factory in Hamburg begins with steel wire, which is cut and pressed into rough balls. These balls are hardened in a series of stoves and then three increasingly precise grinds are enforced until they are spherical up to within a tenth of a micron. The result is one of the most versatile components of modern industry, which enables factors with low friction in everything possible, from speeds to automotors.

This accuracy level requires constant tests – but if defects appear, it can be a mystery. Testing could show a defect that occurs at some point on the assembly line, but the cause may not be obvious. Perhaps the torque of a screw tool is switched off, or a newly replaced grinding wheel has an impact on the quality. Finding the problem means that data is compared across several industrial devices, of which nobody was designed.

This could also be a job for machines soon. Last year Schaeffler was one of the first users of Microsoft’s Factory Operations Agent, a new product that was operated by large -scaling models and specially developed for manufacturers. The chat bot style tool can help to track down the causes of defects, downtime or excess energy consumption. The result is something like Chatgpt for factories, whereby the models from Openai use Microsoft’s Azure in the backend thanks to the company’s partnership.

Kathleen Mitford, Corporate Vice President of Microsoft for Global Industry Marketing, describes the project as “an argumentation agent that works through the manufacturing data”. As a result, Mitford says: “The agent can understand questions and translate it with precision and accuracy against standardized data models.” A factory worker could therefore have a question such as “What causes a higher level of defects?” Place. And the model could answer with data from the entire manufacturing process.

The agent is integrated deep into the existing Enterprise products from Microsoft, especially in Microsoft Fabric, the data analysis system. This means that Schaeffler, which operates hundreds of plants on the Microsoft system, can train its agents on data from all over the world.

Stefan Soutschek, Schaeffler’s Vice President, who is responsible for this, says that the scope of the data analysis is the actual power of the system. “The main advantage is not the chatbot itself, even though it helps,” he says. “It is the combination of this OT [operational technology] Data platform in the backend and the chat bot that relies on this data. “

Despite the name, this is not the agent AI: it has no goals, and its powers are limited to answering all the questions the user asks. You can set up the agent to carry out basic commands via the Microsoft Copilot Studio. However, the goal is not that the agent makes its own decisions. This is primarily AI as a data access tool.