AI agents in logistics: 7 thoughts on the future of the supply chain
By Andre Kranke I 8 minute read
30/07/2026
Agent-based systems in artificial intelligence—hardly any other technology trend is currently the subject of such intense discussion. These seven statements identify what’s important for the successful use of AI agents in logistics in the future.
The logistics industry sees great potential in these digital helpers. As is often the case with a much-hyped topic, lots of people are talking about it without really knowing exactly what it is. The definition of agentic AI varies somewhat from case to case. AI agents are basically software applications. Unlike conventional programs, however, AI agents combine generative artificial intelligence with rule-based computing. This combination results in digital entities that, within a defined scope of tasks, can understand requests and data inputs, derive implementation strategies based on predefined goals, and ultimately trigger and monitor actions.
For example, in shipment tracking, an AI agent can understand information from status systems or even unstructured chat messages, draw conclusions based on predefined objectives, and trigger appropriate actions. When a shipment reaches a certain status, the agent either notifies the customer in question directly or forwards the case to a customer service representative for review and processing.
Statement #1: The lines between AI agents and AI assistants are becoming increasingly blurred
In an ideal situation, an AI agent acts as a fully autonomous system that decides for itself whether and in what way a human should still be consulted. In practice, however, it’s currently not uncommon for an AI agent to function more as an assistant, carrying out virtually no actions without human approval. Depending on the quality of the AI results and process specifications, hybrid forms of AI agents and AI assistants are materializing. The formal boundaries separating the definitions of an AI agent and an AI assistant are increasingly blurred, and the two are thus merging.
“Because of the inherent error rate in AI agents, it’s always necessary to conduct a risk assessment.”
Statement #2: AI agents are transforming logistics processes
The first agent-based systems are being tested, particularly at the interfaces between the various actors along the supply chain. One example is having AI agents enter the necessary shipment data into logistics platforms. This can include unstructured text data and documents, as well as voice messages, and can come from people or other AI agents. Thanks to the new Model Context Protocol (MCP), large language models can communicate more and more effectively with third-party systems and databases, leading to a new form of today’s EDI- and API-based data interfaces. In purchasing and sales, AI agents gather initial information on the characteristics and availability of products and services, review contract terms, and negotiate them. This is already happening today in initial pilot projects.
As part of the “Future Lab” series, results from the Corporate Research & Development department are presented, which were developed in collaboration with specialist departments and branches as well as the DACHSER Enterprise Lab at the Fraunhofer IML and other research and technology partners.
Statement #3: AI agents make mistakes
An AI agent doesn’t always draw the right conclusions and thus can’t always trigger the desired action. This is a fundamental challenge in many operational scenarios. Even perfect AI agents currently can’t exceed an estimated accuracy rate of 90 to 95 percent. Another important factor in this context is the agentic system’s ability to recognize how good its own decision is, what process is at work, and which decision-making rules apply. This capability further influences the quality of an AI agent’s actions. Because of this inherent error rate in AI agents, it’s always necessary to conduct a risk assessment that asks: What impact do occasional errors have, and how does the business model deal with them? It follows that, at present, there are many processes that can’t be entrusted to AI agents.
“In the future, it will be not only people but also AI agents who will shape the future of logistics and supply chain management.”
Statement #4: AI agents are part of the team
AI agents have strengths and weaknesses that are sometimes strikingly similar to those of real people in everyday working life. Together, these qualities make the agents much more than software programs; they’re more like members of the team. They are prepped and trained for various tasks and are capable of interacting with humans. In the future, people will assign tasks to AI agents and work iteratively with them toward the desired result. And not with just a single AI agent; several have to run in parallel, because AI agents need time to “think.” For AI agents to perform their tasks effectively, they must be repeatedly provided with the context of their actions through high-quality data and instructions (prompts). In other words, they need to be managed professionally and integrated well into human teams so that they are accepted as “colleagues.” These are new skills that leaders and managers at all levels must learn very quickly and put into practice in the future.
Artificial intelligence boosts efficiency and supports processes, but the final decision remains with humans. CEO Burkhard Eling explains why this principle is so important to Dachser.
Statement #5: Anyone can create AI agents
Neben dem Führen und Nutzen rückt auch das Erstellen von zumindest einfachen Agenten in den Mittelpunkt des Arbeitsalltags von Logistikern. Unter dem Schlagwort „Citizen Development“ entstehen Umgebungen, mit Hilfe derer viele Nutzer einfache KI-Agenten durch gezieltes Prompten sowie die Anbindung von Datenquellen und Workflow-Systemen erstellen und zum Einsatz bringen können. Ein Beispiel im Büroalltag sind die Copilot-Agenten von Microsoft.
Statement #6: AI agents don’t yet deliver any ROI
Despite the considerable potential of agent-based systems and the initial concrete efficiency gains they achieve in certain processes, they don’t often yet yield a return on investment. Many AI agents are still at the minimum viable product (MVP) stage, are currently being tested in pilot projects, and are still far from being fully integrated into business models. Development costs, licenses for AI systems, and token consumption for queries to large language models such as ChatGPT, Claude, or Mistral all come at a cost—especially if there’s an overreliance on a single provider or LLM. Added to this are the training sessions for employees and the time needed to gain experience in working with AI agents. Every innovation process of course involves a certain amount of investment in learning how to use new technologies properly. But step by step, specific KPIs and ROI analyses must also become a guiding tool for AI agents in order to create a sound economic framework for the use of artificial intelligence in the medium term.
Statement #7: AI agents are shaping the future
Despite their weaknesses, risks, and costs, AI agents will make inroads into nearly every process in the supply chain. Everyone should be aware of this. AI agents will sometimes change the nature of collaboration only slightly, but in some cases quite significantly, and in most cases they will make it more effective. This involves not only individual AI agents, but also multi-agent systems that interact with one another and with humans.
In the future, therefore, it will be not only people but also AI agents who will shape the future of logistics and supply chain management. This means that working with AI agents will also become a leadership responsibility. Stefan Hohm, Chief Development Officer (CDO) at DACHSER, sums it up: “AI will not replace leaders. But leaders who use AI will replace those who don’t.”
