Pilot Project: Pharmaceutical Giant Relies on Artificial Intelligence from Styria
Around 100 million medicinal products leave the Graz facility of the international healthcare company Fresenius Kabi each year. The processes for batch production are analyzed and documented according to the strictest national and international standards: For this purpose, artificial intelligence (AI) is now being utilized in a pilot project. The implementation is the responsibility of the Graz-based AI pioneer Leftshift One.
- 17. October. 2024
Karin Schnedlitz
Content Managerin
Medicinal products such as anesthetics, analgesics, and antibiotics leave the Fresenius Kabi facility in Graz in the millions every year. In the production of these predominantly liquid medications, the highest national and international quality standards are paramount: “Since our products are destined for the national and international pharmaceutical markets, millions of lives around the globe depend on their quality. These products must be physically, chemically, and microbiologically perfect and sterile. No deviations are tolerated,” emphasizes plant manager Ruth Staubmann regarding the stringent quality regulations at Fresenius Kabi Austria. Therefore, the analysis and documentation of every conceivable occurrence on the production lines hold particularly high importance: “Production is an industrial process involving both people and machines under the most demanding requirements. In rare cases, deviations from predefined processes can occur. What matters then is assessing the impact of the deviation from the plan on the product,” explains Christoph Köth, Head of Innovation at Fresenius Kabi Austria. This means that even minimal changes to the planned process are analyzed and evaluated according to precise guidelines—and, in extreme cases, may even lead to the destruction of a batch, according to the innovation leader. The key factor is that the faster, more accurate, and clearer the documentation of deviations occurs, the more targeted the corresponding actions can be initiated.
"Our technology is capable of analyzing and categorizing the textual descriptions of employees. This transforms previously unstructured texts into comparable datasets."
Patrick Ratheiser, CEO Leftshift One
Premiere for Artificial Intelligence
Together with technology specialist Leftshift One, Fresenius Kabi Austria is implementing artificial intelligence on its production lines. “With the help of AI, we want to ensure that the information for analyzing a deviation is sufficient and complete—regardless of parameters such as the documenting employee or the time of day,” explains Köth. This means that while previously the documentation by the production employee was solely interpreted by quality experts, AI now also performs a preliminary classification of the incident. “Our technology is capable of analyzing and categorizing the textual descriptions provided by the employee. This transforms previously unstructured texts into comparable datasets,” explains Patrick Ratheiser, CEO of Leftshift One. Past events on the production lines are integrated into the AI’s preliminary classification. Based on AIOS, the operating system for artificial intelligence developed by Leftshift One, the Graz-based company can implement such AI applications quickly and reliably. “With our technology, we can offer initial solutions to companies across various industries within a few weeks. It is essential that the respective domain knowledge is linked with our data expertise—and that humans remain at the center of this process. This is particularly important in this project, as the AI’s preliminary classification must be understandable, transparent, and explainable for the user,” emphasizes Ratheiser. For the CEO, only the symbiosis of humans and machines makes the application of AI meaningful.
"The better the recommendations become, the higher the acceptance among people. Therefore, we must ensure that the employee proactively verifies the suggestion from the artificial intelligence at all times."
Christoph Köth, Innovationsleiter Fresenius Kabi
Thus, at Fresenius Kabi Austria, the final decision on how to handle deviations from the plan still rests with the quality experts. “It is important that we provide an impartial quality assistant with this AI-based tool during the decision-making process,” emphasizes Innovation Manager Köth. In this context, the healthcare company is examining through thesis projects to what extent it is necessary to implement mechanisms to ensure the independence of the quality experts’ decisions from the AI’s suggestions. “The better the recommendations become, the higher the acceptance among people. Therefore, we must ensure that the employee proactively verifies the suggestion from the artificial intelligence at all times,” says Köth.