AI language models understand customer needs

Dresden-based software provider LOGSOL uses Artificial Intelligence to create individual process solutions faster and more efficiently. The company is developing its expertise in close collaboration with TU Dresden

“The increased expectations of availability and functionality of software solutions in logistics” are the biggest challenges for David Wustmann in his working environment. “In the best case scenario, it should cost nothing,” says the Head of Logistics Software at the cloud solution provider LOGSOL. Customers and users include logistics service providers in transportation and warehousing.

The medium-sized company with more than 120 employees at eight German locations tries to “occupy niches with very extensive expertise” and works together with universities and research institutes throughout Germany. At the company’s headquarters in Dresden, for example, there are collaborations with the Technical University and the University of Applied Sciences, as well as projects with the University of Siegen, the Fraunhofer Institute for Material Flow and Logistics in Dortmund as well as start-ups, among others.

LOGSOL has been using Artificial Intelligence (AI) in product development for two years in order to write software codes more efficiently. Wustmann himself completed his doctorate at TU Dresden on artificial neural networks, which are used for machine learning and AI. The business information scientist and mechanical engineer explains how Microsoft’s AI assistant Copilot, for example, supports software developers with documentation:

“It points out security leaks in the code directly or recommends writing an automated test.” The team also learns how to “work efficiently with AI technology”, he says.

Effective service with AI

In future, LOGSOL would like to offer features in its own container management or time slot management systems for intralogistics and along the supply chain so that customers can also benefit from AI technology. Wustmann explains: “Without AI, it would not be possible to effectively answer questions about the customer-specific process by using a manual or a standard chatbot, for example.”
The Dresden-based company does not sell standard software. The idea is to implement a support AI and train it with specification data according to the customer’s requirements documents. In future, AI will answer a question about a specific process for which the provider’s software experts would have to work through hundreds of pages of application specifications.

This relieves the support department enormously and enables more favorable conditions, which is also interesting for logistics service providers. Manual bookings in the system – scanning, entering and typing in data – are no longer necessary. Wustmann also sees this as a way of countering the shortage of skilled workers and “giving employees meaningful other activities”. The software developers are also working on using AI to improve specific work processes in cross-company collaboration between logistics service providers and commercial or industrial companies. One example: For supply chain planning, owners of returnable load carriers want to track where the containers are located and whether they are available.

This works on the basis of bookings or via a GPS signal or mobile network location in real time. “However, tracking technologies are only worthwhile for very expensive special load carriers, not for the many pallet cages and small load carriers such as plastic or thermo boxes,” explains Wustmann. Tracking technology is therefore significantly more expensive than the value of the container.
Together with TU Dresden and tracking manufacturers, LOGSOL is investigating how costs can be saved through intelligent digitalization in the supply chain.

For example, whether it would be practical to track shipments with just one tracked container per truck that transports the goods between the supplier and the production company. Statistical evaluations and machine learning are used for this. Approximate values would be sufficient for this – and according to Wustmann, AI can solve this task very well. It would have to learn how delivery flows change in order to maintain a robust level of accuracy without changing the number of containers equipped with tracking technology.

Knowledge transfer from university of excellence

The current project is funded by the European Regional Development Fund (ERDF). TU Dresden is one of the ten universities of excellence in Germany selected on the basis of the excellence strategy of the Federal Ministry of Education and Research and is participating in the project on an interdisciplinary basis with one computer scientist and one logistics specialist.

“The transfer from the academic side is very important to us,” emphasizes Wustmann. It was preceded by the master’s thesis of a working student in the Business Mathematics program at the TU Bergakademie Freiberg, who worked intensively on mathematical models and simulations to predict container movements.
According to Wustmann, five working students are currently active for the company, contributing new ideas, such as process engineering for the efficient loading of trucks. Another project that could reduce the shortage of skilled workers is currently being transferred to the ERDF. The idea is to employ people who have changed careers in the warehouse or as truck drivers, even though they lack language and/or specialist knowledge. AI could, for example, help employees working in incoming goods to operate an input screen correctly – even if they do not understand the underlying process.

Another application example is time management at a freight forwarding company, where ChatGPT could act as a language bot to assist Arabic- or Italian-speaking employees in their native language and finalize bookings. All these requirements are practice-driven because EU Directive 2019/882 requires LOGSOL customers to check their software for accessibility from 2025 and adapt it if necessary.
Wustmann recommends that medium-sized logistics service providers “start using the new tools today”. In his view, recurring processes or activities such as processing, checking and approving invoices can be automated very well in the first instance – “ with AI or without”. (loe)

Editorial article published on July 23, 2024 in the DVZ (Deutsche Verkehrs-Zeitung) | Kerstin Kloss