Dimitri De Rocker and Simon Uytterhoeven (Datashift): “Generative AI opens up new opportunities for logistics processes”
Over the past two years, insights about Artificial Intelligence (AI) have changed fundamentally. Generative AI suddenly opens doors that machine learning – the previous generation of AI – could only leave ajar. Data, the raw material of AI, becomes even more crucial as a result. Dimitri De Rocker, AI director at Datashift, and Simon Uytterhoeven, ‘AI Translator’, share their insights on how generative AI will transform processes in logistics.
AI and the flow of data in logistics
Logistics is a sector in which many documents circulate, often from external parties. Administrative teams have the challenge of extracting relevant information from these documents and processing it efficiently. Moreover, some data, often collected manually, must first be digitised to be usable in Warehouse Management Systems (WMS) or Transport Management Systems (TMS). For optimal cooperation between these systems, high data quality is essential. AI-driven automation helps here by making processes more efficient and reducing human errors.
The role of data in AI applications
“To properly understand AI applications, we need to distinguish between structured and unstructured data. As is well known, data fuels AI, but the type of data determines which AI model is suitable and which applications are possible,” says Simon
“Structured data is data that can be organised into tables and easily analysed. AI models can process very large amounts of data to calculate transport costs, optimise maintenance schedules or increase operational efficiency, for example. Machine learning can process large amounts of structured data and identify trends better and faster than a human, with human intervention only needed for exceptions,” he explains.
“With AI, it is relatively easy to extract insights from this data, such as which transports were profitable and which were not. With large amounts of information, this becomes cluttered for humans, whereas AI can recognise patterns and forward this information to a decision maker, or to other operational systems, such as a CRM platform,” Dimitri adds.
Unstructured data: the new challenge
Equally great potential lies in unstructured data, from emails, chats, PDF documents and even images. Until recently, it was difficult to exploit this data efficiently, but generative AI is changes this situation. The models that form the basis of tools like ChatGPT, Copilot and DeepSeek – which have been on the market for less than three years – can structure unstructured information so that it is turned into usable data for operational systems.
Simon: “Generative AI can automatically analyse documents and extract relevant data. Think notes of lading, invoices or customs documents that are processed and loaded directly into logistics systems.”
“In addition, unstructured data can also be directly exploited. An example is analysing video footage at a container terminal. Artificial Intelligence can automatically detect damage to containers and determine if historical image data exists – whether it has occurred recently. Image recognition can also be used to check compliance with safety regulations, such as the wearing of helmets or safety jackets, or still monitoring forklift movements, for instance to determine whether they cross certain lines. This allows companies to take measures that improve safety and efficiency,” Dimitri adds.
Generative AI: a new phase for logistics
Artificial Intelligence has suddenly become tangible to the general public with tools like ChatGPT. Although AI technology had been around for decades, it was often less visible. Think of the algorithms behind streaming recommendations or personalised ads. Recent developments are making AI more accessible and versatile.
Dimitri explains: “Traditional AI models, based on machine learning, mainly focus on analysing data to make predictions, or divide it into similar groups. Machine learning models are trained on structured data and improve themselves as they process more examples, often under human supervision. They do not generate genuine new information”.
“Generative AI, on the other hand, works differently. Generative AI is pre-trained on large amounts of data and will generate new output based on an external input given to the model. You provide a context, the model appears to understand that context, and thus creates a new text, image or analysis that did not exist before. This makes the technology particularly powerful for process optimisation.”
Simon added: “With generative AI and smart automation, we can process and structure information from different fields and from different origins and layouts, which fits processes together seamlessly. As a result, AI can not only support administrative tasks, but even take them over completely. If the AI system is not sure of the data, it can be checked via already validated information (e.g. master data) or by an employee. The latter we call a ‘human-in-the-loop’.”
The impact on logistics is significant. Whereas previous AI models still required complex implementations, generative AI is much more accessible and widely deployable. This results in a faster return on investment (ROI), often within as little as one to two years.
About Datashift
Many companies are not yet leveraging their data sufficiently or do not know how to use it effectively. Others have data, but lack insight into possible applications or the importance of a robust data platform.
Datashift, a Belgian data consultancy, helps organisations turn their data into valuable insights through strategic data analysis, engineering and governance. AI plays a very important role in this development.
As theme partner of ‘The Future of Work’ at LogiVille, Datashift will demonstrate how AI can process unstructured data, such as documents and video footage, to optimise logistics processes. Examples include extracting relevant data from PDF documents, automatic damage inspections or inventory management via image recognition.
