DHL maps agentic AI shift in logistics

DHL has placed agentic artificial intelligence among the technologies expected to exert the greatest influence on logistics over the next decade, as its latest Logistics Trend Radar tracks AI moving from analysis and assistance towards planning, decision making, and operational action.

The eighth edition adds agentic AI, AI commerce, the compute economy, and humanoids to DHL’s trend map. They sit alongside areas including AI analytics, generative AI, augmented workplaces, cybersecurity, supply chain diversification, sustainable fuels, and decarbonisation, producing a picture in which digital systems and physical logistics are increasingly connected.

Agentic AI marks an important change in the way software can participate in an operation. Earlier AI applications have largely searched information, generated content, identified patterns, or produced recommendations for a person to consider. Agentic systems are intended to work towards defined goals and complete more of the sequence themselves.

Potential applications include identifying inventory shortages, rerouting shipments around disruption, reallocating transport capacity, and coordinating information across several planning and execution systems. DHL also identifies AI analytics as an important companion technology, allowing operators to detect patterns and forecast events before an agent or employee decides how to respond.

Klaus Dohrmann, Vice President of Innovation and Trend Research at DHL Customer Solutions & Innovation, said: “The next chapter is about action.” Moving from recommendation to execution changes the operational risk considerably because an incorrect automated decision can affect inventory, capacity, customer commitments, or physical material flows before a planner intervenes.

That puts greater emphasis on the quality of the information underneath the AI system. Logistics organisations rarely operate with one perfectly standard data environment. Warehouse management platforms, transport systems, customer applications, sensors, spreadsheets, enterprise software, and older local systems can all describe the same operation in different ways.

Agentic software can act quickly on that information, but speed is of limited value when source data are inconsistent or incomplete. Permissions also become more consequential. A system authorised to prepare an analysis carries a different level of operational risk from one allowed to alter an allocation, contact a carrier, release a workflow, or change a shipment plan.

The shift is already visible in operating warehouses. CJ Logistics America has expanded agentic AI across more than 40 facilities, connecting multiple warehouse management systems with enterprise data, sensors, and automation equipment. Agents are being used for areas including labour performance, compliance, gap time, travel, and slotting rather than remaining confined to a demonstration environment.

That deployment provides a useful contrast with the Trend Radar. DHL is identifying the direction of travel across the sector, while individual logistics operators are beginning to establish where autonomy delivers measurable value and where decisions still need direct human control.

Workforce technology occupies a substantial part of the new Radar for the same reason. Wearable sensors, remote operations, collaborative robotics, and AI workplace assistants can reduce repetitive work and accelerate access to information, but they also alter the skills required from supervisors, engineers, planners, and frontline staff.

As software takes responsibility for routine analysis or execution, employees increasingly deal with exceptions: unusual orders, failed equipment, inconsistent data, customer changes, safety events, and situations outside the automation rules. Those tasks demand judgement and an understanding of how several systems interact, rather than simply following a standard transaction sequence.

Governance consequently becomes an operating discipline. DHL continues to track AI ethics and Cybersecurity 2.0, reflecting the additional exposure created when more systems can act on commercially or operationally sensitive data. Access controls, audit trails, recovery procedures, and clear boundaries around automated authority become part of logistics system design.

Humanoids take the discussion into physical automation. Conventional warehouse robots perform well where routes, loads, or movements are tightly defined. Humanoid systems are being developed for environments already designed around human reach and movement, potentially reducing the need to rebuild a workstation around every automated task.

Commercial maturity remains uneven. Automated storage, conveyors, autonomous mobile robots, and robotic pallet handling already operate at scale, while general purpose humanoid systems remain an emerging category. Their inclusion in the Radar signals where research and investment are moving rather than evidence that warehouses are about to replace established automation fleets.

The compute economy introduces another physical consequence of AI growth. Data centres require servers, electrical equipment, cooling technology, construction materials, replacement components, and specialist transport, so increasing digital capacity creates sizeable logistics requirements of its own. AI therefore adds demand to the infrastructure that supports logistics even as it is applied to logistics operations.

Sustainability continues alongside those digital trends. Heavy vehicle electrification, sustainable fuels, circularity, and decarbonisation remain on DHL’s map, meaning operators are likely to manage several overlapping transitions rather than moving cleanly from one technology cycle to the next.

The more difficult task will be integration. AI, robotics, fleet electrification, cyber controls, warehouse systems, and workforce tools all compete for data, technical expertise, investment, and infrastructure. The useful deployments will be the ones that remove operating friction without creating another layer of systems that employees have to reconcile manually.

DHL’s eighth Trend Radar does not demonstrate that autonomous supply chains have arrived. It identifies a more specific shift: software that previously explained operations is increasingly being given permission to participate in them. The commercial test now moves from what the models can generate to what logistics businesses can safely allow them to do.