- Special Logistics
- Blog
Unleashing the Power of GPT in Logistics
Can we imagine our future without AI-based assistants?
We have probably all interacted with ChatGPT, the AI chatbot that has captured the world’s attention. But what exactly is ChatGPT, and how can DB SCHENKER leverage its power to enhance business operations?
Understanding the Foundation
Generative AI is a state-of-the-art technology capable of producing a wide range of creative content, from text and images to music and video. Large Language Models (LLMs) are a subset of Generative AI specifically designed to generate text based on prompts and conversations. It was trained on a large set of textual data from the web and designed to perform chat-like conversations with humans on a broad range of topics based on its large knowledge base. ChatGPT excels beyond simple question-answering. It retains conversational context, enabling it to ask follow-up questions and provide tailored responses, making it more than just a question-answering machine. Looking at Generative AI, DB SCHENKER is diving deep into the world of GPT. Let’s uncover the potential of GPT in transforming the logistics industry.
Potential Applications in Logistics
GPT is more than just a Chatbot. GPT’s exceptional natural language understanding, text processing, and conversational abilities make it an intriguing tool for DB SCHENKER. These skills are being leveraged to explore practical applications within daily logistics operations.
Automated Data Extraction
The logistics industry generates a massive volume of documents daily, from invoices and delivery notes to customs declarations. Manual processing of these documents is time-consuming and prone to errors. GPT, in conjunction with Optical Character Recognition (OCR) technology, can automate the extraction of information from these documents, significantly reducing manual effort and accelerating processing times. By combining GPT with OCR technologies, DB SCHENKER can process and understand texts within documents, structuring the information in a way that seamlessly integrates it automatically into DB SCHENKER systems. This automation is being tested across various document types, including customs files, invoices, master bills of lading, export documents, and claims.
Intelligent Information Assistant
The DB SCHENKER global organization stores a large document base such as manuals of IT systems, Standard Operating Procedure (SOP) documents, guidelines, and training materials. Finding information in this large data set is a time-consuming and nerve-wracking task. GPT can serve as an intelligent information assistant, quickly searching through vast amounts of data to provide relevant answers to user queries. This approach tested at DB SCHENKER improved the onboarding to new systems, and understanding of procedures, therefore saving time, and making work more efficient. At DB SCHENKER, the Intelligent Information Assistant approach is especially used to find information in SOPs, IT system manuals (e.g, TANGO), HR policies, FAQ/Q&A topics, or scientific literature.
Data Cleansing and Correction
Data quality is crucial in logistics. Inconsistent data formats, errors, and duplicates can hinder decision-making and operational efficiency. Some of the data at DB SCHENKER is inherently incorrect due to potential human error and a variety of different IT systems. There are various ways to write down the same address or the Master Data contains duplications or wrong entries. GPT can cleanse and standardize data, correct address details, identify and eliminate duplicates, and flag potential errors. At DB SCHENKER, the primary focus is on leveraging GPT for tasks such as correcting addresses for transportation, cleansing Master Data, de-duplicating data sources, and data validating and comparing data.
Navigating the Challenges
All AI solutions have limitations. It's important to note that human oversight and intervention are necessary to ensure that the content generated by GPT is legal, accurate, and relevant. Ownership of the generated content is unregulated. This needs to be clarified to avoid infringement of third-party intellectual property. The AI Act is supposed to clarify and provide guidance on this topic in the European Union.
Data privacy is a paramount concern in AI applications. The protection of personal data should be also handled in an AI system. There is an ongoing debate about data collection and privacy, which is a subject to be clarified with the LLM/GPT vendor. To mitigate risks, it is highly recommended not to feed personal data into AI solutions even if all data remains within the DB SCHENKER network.
Evaluation and Considerations
While initial trials of GPT applications at DB SCHENKER show promise, it is important to note that these are early-stage results and do not validate a large-scale, global use of the solution. Validating a few selected examples may not provide a comprehensive understanding of the technology’s capabilities and limitations. Overfitting to specific use cases can hinder its broader applicability. Before scaling GPT applications, an evaluation process is crucial. By measuring errors, collecting representative data samples, and assessing the potential business impact of inaccuracies, DB SCHENKER can ensure the development of high-quality GPT solutions that meet user expectations and business needs.
