From Warehouses to Last Mile, AI Is Rewiring Logistics at Global Scale

For DHL Supply Chain, the world’s largest contract logistics provider, logistics has become a software problem as much as a physical one. The company now operates across 220 countries, runs more than 7,500 autonomous warehouse robots, and relies on machine-learning systems to predict inventory discrepancies, labor risk and delivery bottlenecks. More than 90% of their warehouses have at least one automated solution.

That infrastructure, DHL executives say, is what allows the company to absorb the volatility of product launches, seasonal peaks and shifting consumer demand without breaking.

Digital Foundation Built Before AI

DHL’s approach to artificial intelligence (AI) did not begin with generative models or autonomous agents. According to Jason Pawlowski, vice president of IT at DHL Supply Chain, the company spent years standardizing processes and cleaning data before deploying advanced automation.

“Innovation that doesn’t scale is just a nice idea,” Pawlowski said during a “CES Tech Talk Inside Innovation” episode, describing why DHL avoids deploying technology for its own sake.

The company first focused on digitizing core warehouse operations and putting consistent performance data into the hands of managers. That work enabled machine-learning systems that now automate tasks once handled manually, including discrepant receiving for returned electronics. Instead of days of back-and-forth with customers to determine how devices should be processed, models predict the correct handling path with high accuracy, cutting time and cost out of the system.

“You can’t get to trustworthy AI without trustworthy data,” Pawlowski said, framing data hygiene as a prerequisite rather than a byproduct of automation.

That same data backbone supports digital twins that allow DHL and its customers to model what-if scenarios across the network, testing how changes in warehouse location, transportation mix or inventory allocation would affect delivery speed and cost before anything moves in the real world.

Boston Dynamics Pushes Warehouse Robotics

Warehouse robotics has moved beyond fixed automation and fenced-off systems, a shift underscored at CES by Boston Dynamics. The company’s new Atlas robots are designed to perform discrete, physically demanding tasks that historically constrained throughput, including wide array of industrial tasks, from material handling to order fulfilment.

DHL Supply Chain has previously partnered with Boston Dynamics to deploy Stretch, a warehouse robot that autonomously unloads trailers and handles cartons. Stretch deployments have shown unloading rates of up to roughly 700 boxes per hour and reduce physical strain for workers.

Neolix Targets the Last-Mile Bottleneck

At CES, Neolix also highlighted how autonomy is extending beyond warehouses into transportation. The company debuted their expanded RoboVan lineup designed specifically for commercial logistics.

Neolix positioned itself as one of the first companies to commercialize Level 4 autonomous driving without relying on high-definition maps. That approach lowers deployment costs and shortens setup times, enabling RoboVans to operate in dense urban conditions, mixed traffic and adverse weather, and making it easier to expand autonomous delivery into new cities and countries without extensive mapping work.

Inside DHL, autonomy is being introduced with similar restraint. Agentic AI systems are first deployed on low-risk, high-volume tasks such as delivery appointment scheduling and urgent order notifications.

“We used to spend days calling customers to schedule delivery appointments,” Pawlowski said. “Now with the use of AI agents, that work can be done in hours.”

Other models predict inventory locations most likely to have discrepancies and identify labor retention risks, prompting earlier intervention by managers. Pawlowski framed these systems as augmenting human judgment rather than replacing it.

Looking ahead, he said the next step will be coordination between autonomous systems across vendors and platforms.

“If these agents can start collaborating across platforms, bots working with bots, that would be orchestration on a whole new level,” Pawlowski said.