Logistics is in the middle of a major transformation: many processes integral to supply chains that were once purely manual are now becoming increasingly automated. It generally takes a while for the same to happen to internal processes. But these also offer considerable potential—to make a company’s internal information management more efficient, for example. Pick any international company and chances are its employees are constantly asking similar questions: “What contract conditions apply to our Swiss customers? What’s our policy for mobile working? What’s the correct procedure for requesting vacation time?” But to find answers to these questions, employees often have to wade through a quagmire of company data, which eats up a great deal of their time.
Michael Lütjann, CIO of Nagel-Group: “Whether it’s in sales, real estate, HR, or IT, our company generates so much information that our employees can no longer process it effectively. As a result, they had no quick way of accessing the information they needed at a given moment.” In the past, when Nagel-Group employees required information relating to a contract or works agreement, they had to search the data management system or the intranet for the specific keyword: “But even with that exact keyword, you might find only some or none of the information you were looking for—or you would end up with far too many results. This was because the static full-text search function wasn’t smart enough to link related pieces of information,” says Sven Kreciszewski, IT Innovation Manager at Nagel-Group. “But that’s precisely what generative AI can do.”
Nagel-Group decided to address this situation early on by entering into an agreement on artificial intelligence (AI) with the works council. “That paved the way for all further AI use cases and our NagelGPT solution based on Azure OpenAI Service. Looking back, it was a crucial decision that enabled us to get to grips with this topic really quickly,” Lütjann says.
The solution: NagelGPT based on Azure OpenAI Service always provides the right answers
Nagel-Group started by working with its partner P3 to prioritize those use cases that offered employees the most added value. “These tend to be the works agreements that apply throughout the company. It’s up to the employer to make sure these are honored,” Kreciszewski says. Now, when employees want to find out what the current rules are governing mobile working or giving Christmas presents to business partners, they just open NagelGPT in Microsoft Teams and enter their question. Behind the scenes, Microsoft Copilot Studio forwards the question to the central application via Power Automate.
Azure OpenAI Service helps give each inquiry its own content “fingerprint” and interprets references to filter parameters. This then forms the basis for starting the search through all documents via Azure AI Search. As soon as the system finds an answer, it gathers all the pertinent information and creates a suitable response comprising the question, text blocks, and chat history. “The response the users receive also includes all sources and references for the original text passages. In addition to shortening the search time, this makes it easier to delve deeper later on,” says Dr. Nils Drescher, Head of AI & Data Analytics at P3. “The semantic search function in NagelGPT provides employees with information that builds on their current level of familiarity with the subject matter. They can now pose much more open questions about a given topic and still receive a satisfactory answer.”
As soon as Nagel-Group had implemented the first round of use cases, it moved on to the second: “Thanks to the system’s modular structure and the powerful models available through Azure OpenAI Service, integrating new functions and documents at short notice is no problem at all. That means we can always respond quickly to inquiries from the different departments and ensure that the data corpus is kept up-to-date,” Kreciszewski says. This is particularly important for the sales team, whose documents are currently being added to NagelGPT. “If the parameters suddenly shift, we have to be able to call up details of the affected customer contracts right away,” Lütjann says. “NagelGPT is a big help here as well. We can also ask our knowledge bot when certain contracts are due to expire and which contract applies to a specific product and at what location. This allows us to put together an overall picture of contract parameters. It helps us increase transparency and standardization because it means we can offer our customers consistent conditions across all locations. And it provides the basis for more predictive planning for transports.”
NagelGPT is fully integrated into the Microsoft environment at Nagel-Group. “The upshot was that we could work within our existing infrastructure without having to first hammer out new agreements on data protection, data availability, and data processing. This, plus the collaborations with Microsoft and P3, guaranteed that progress would be swift.” It was also really important to Nagel-Group employees that the solution be integrated into existing tools like Microsoft Teams: “Giving AI the right questions, or prompts, is an art in itself and we have to support our colleagues by showing them how this works. If they were having to grapple with a new tool as well, they could get really frustrated really fast,” Kreciszewski says. “But with help from P3, we managed to navigate these transformation topics with ease.”
This experience also sets a clear course for the future: Nagel-Group wants to continue pooling expertise in the areas of industry knowledge, technology, and implementation. “Through the introduction of Copilot for Microsoft 365 as part of an early adopter program, we will achieve further efficiency gains, set new trends, and drive the digital transformation of logistics forward. It was clear from the very beginning that AI would massively change the industry. That’s why we knew we had to get a head start on this topic so we could set the tone and help shape the AI transformation.”
Voices on the NagelGPT
