elunic democratizes AI for industry with shopfloorGPT, an autonomous problem solver based on Azure OpenAI Service

Checking performance, managing knowledge or requirements—in the life of a machine, a lot of information and data needs to be documented and evaluated. The software company elunic has set itself the task of supporting small and medium-sized industrial companies and machinery manufacturers in particular on their digital transformation toward Industry 4.0. To this end, elunic has expanded its shopfloor.io platform to include shopfloorGPT. Based on Azure OpenAI Service, shopfloorGPT functions for elunic’s customers as a digital full-time employee tasked with supporting quality inspections, service requests, and production monitoring.

The challenge: Taking manufacturing to the next level

Imagine a machine that not only collects its own telemetry data—making it more transparent and optimizable—but also regulates its own operation based on this data. If it runs too hot, it slows itself down automatically. And if an error occurs, it sounds the alarm and immediately provides the technical staff with a suggested solution. In short, it’s an autonomous machine. And elunic has set itself the goal of helping to shape precisely this vision. “For us, autonomous machines are the guiding star that spurs us on to be more innovative and helps our customers stay one step ahead of the competition. The journey often begins with optimizing in-house services through the use of AI agents,” says Rebecca Reuter, Senior Marketing Manager at elunic GmbH.

But there’s still a long way to go before this vision of the future is realized. Although machines already deliver several terabytes of data every day, the potential within that data isn’t yet being exploited. “Some of our customers are sitting on huge troves of data they’re not using, simply because they lack the capacity and expertise,” explains Claudio Gusmini, Executive Director at elunic. And this is despite the fact that the data is relevant to a wide range of areas—for example, quality inspection or production reports. In addition, machinery manufacturers often still carry out data evaluations manually, which takes time; employees have to tag the relevant data by hand and prepare it for subsequent use.

elunic developedshopfloor.io, its digital machine portal, with precisely this purpose in mind. The platform collects data and connects machines, processes, and products, which makes it the first step toward autonomous machines. “However, we wanted to find a solution for our customers that could evaluate data even more efficiently,” Gusmini says. It soon became clear that this called for innovative technologies such as large language models (LLMs). “Together with a lead customer, we then tested the possibilities of AI in mechanical engineering and very quickly saw some initial successes,” Gusmini is delighted to report. The result of these tests wasshopfloorGPT, which is based on Azure OpenAI Service.

The solution: Customizable AI agents based on Azure OpenAI Service

Fast, always available, flexible, and reliable: these qualities are what sets shopfloorGPT apart. Various agents are available within the solution, and each is specialized in a different area of application: service, sales, maintenance, and projects. However, elunic customers can also create an agent of their own, tailored to their individual needs. Available for queries 24/7, these full-time AI employees access various data sources across the company—including databases, machine data, manuals, and in-house SharePoint pages—and provide the requested information within seconds. “For example, the Service Agent saves our customers around 15 minutes per processed request—that’s a huge productivity boost,” Gusmini says.

It’s all thanks to Azure OpenAI Service. The Microsoft technology enables elunic to always use the latest version of ChatGPT as a cloud-based foundation for its agents—securely and in compliance with data protection regulations.Azure Kubernetes Service supports the containerization of the data. This infrastructure ensures that the AI agents can be integrated into all existing environments—from the shopfloor.io machine portal to SAP systems or Microsoft applications and other environments.

“The way shopfloorGPT works is completely different than conventional AI chatbots. Our AI agents are designed to constantly think like autonomous problem solvers, and they have access to a comprehensive set of tools to help them,” explains Konrad Pustka, Director IoT Products at elunic. When an AI agent receives a request, the first step is to make a plan of what information it needs to call up and when, share that plan in the chat, and execute it. In subsequent runs, the agent analyzes its answers and the feedback from users to improve itself. But users can also actively influence the AI agent and help it learn; using a skills function, they can define abilities and requirements and make these available to the agent.

Another helpful function is that shopfloorGPT can automatically create routines, or defined sequences of instructions for executing a task, based on chat histories. For example, say a production employee regularly needs information from the AI agent about a particular machine’s current production quality. She creates a routine based on the chat history so she doesn’t have to enter the same request every time. The generative AI recognizes what information it needs to output based on the information entered and the feedback. It then automatically sends the employee the report she needs straight to her email inbox. This creates a central knowledge database that maintains itself independently and safeguards the know-how of elunic customers. “With our AI Professional Solutions and the use of AI agents based on Azure OpenAI Service, we’re democratizing knowledge and closing process gaps in mechanical engineering. This means we’re doing a great deal to increase competitiveness and resilience,” Gusmini says.

Microsoft supplies more than just the technology for shopfloorGPT. “Despite our years of experience in setting up digital solutions for machines, for this project we decided to bring in AI experts from Microsoft via the FastTrack program to help us take an even more targeted approach,” Gusmini explains. And that’s not about to change anytime soon, because the vision of autonomous machines is clear, but it’s still a long way off. The first step toward realizing it has now been taken in the form of AI agents that help machinery and plant manufacturers tap and leverage their troves of big data. But on the path to the factory of the future, many more steps will follow.