Basic machine For grinding a steel ball bearing has been the same since around 1900, but manufacturers constantly automate everything around. Today, the process is powered by a belt conveyor and, for the most part, is automatic. The most thinner task for people is to determine when things go wrong – and even this can soon be conveyed to AI.
This level of precision requires continuous testing – but when flaws appear, tracking them can present a puzzle. Testing can show the defect at some point on the assembly line, but the cause may not be obvious. Perhaps the torque on the screw tool is turned off or the newly replaced grinding wheel affects the quality. Tracking the problem means comparing data on many elements of industrial equipment, none of which has been designed with this in mind.
This can also be a task for machines. Last year, Schaeffler became one of the first users of the Microsoft Factory Operations agent, a novel product powered by vast language models and designed specifically for manufacturers. The chatbot -style tool can facilitate to track the causes of defects, downtime or excessive energy consumption. The result is something like chatgpt for factories, and the OpenAI models are used in the back room thanks to the company’s cooperation with Azure Microsoft.
Kathleen Mitford, Microsoft Vice President of the Microsoft for global industry marketing, describes the project as “an reasoning agent who works in addition to production data”. As a result, Mitford says: “The agent is able to understand questions and explain them with precision and accuracy in relation to normalized data models.” So a factory employee may ask the question “what causes a higher level of disadvantages than usual?” The model would be able to respond to data from the entire production process.
The agent is deeply integrated with existing Microsoft Enterprise products, especially Microsoft Fabric, his data analysis system. This means that Schaeffler, which supports hundreds of plants in the Microsoft system, is able to train his agent in data from around the world.
Stefan Soutschek, vice president of Schaeffler, responsible for this, claims that the scope of data analysis is the real strength of the system. “Chatbot itself is not the main benefit, although it helps,” he says. “This is a combination of this OT [operational technology] Data platform in backend and chatbot consisting in this data. “
Despite the name, this is not agent artificial intelligence: there are no goals, and its rights are narrow to answering all questions that the user asks. You can configure an agent to follow the basic commands via Copilot Studio Microsoft, but the goal is for the agent to make his own decisions. It is primarily AI as a data access tool.