Interface Standardization Is The Key To Scaling Robotics And AI In Warehouses, Says DHL Supply Chain VP

Warehouse technology has moved from tidy WMS and TMS deployments into a messier world of robotics, agentic AI, and operationally sensitive automation. That jump in complexity has turned integration from back-end plumbing into the make-or-break layer of the modern warehouse stack. When the interfaces are standardized, the system can absorb new tools without buckling. When they’re treated as one-off project work, every new robot, workflow, and data feed becomes another place for operations to crack under load.

Tim Bolam is Vice President of Project Delivery at DHL Supply Chain, the contract logistics arm of Deutsche Post DHL Group. He has delivered more than 7,600 programs globally across retail, FMCG, automotive, healthcare, and e-commerce, and leads the standardization of integration, implementation, and operational playbooks across DHL's warehouse technology deployments.

"I'm still amazed at how many customers don't have standard integration and how many manual bits there are around that process," said Bolam. "A lot of customers say they only want five or six interfaces, and then when you start talking to the people who really operate the business, you find more and more levels of integration." Bolam's core argument is that standard-first integration is the only way for large operators to absorb the complexity now built into every deployment, and that the organizations still treating integration as temporary plumbing are building systems too fragile to scale.

• Standard messaging at the start: DHL operates two primary WMS platforms, Manhattan Active for e-commerce and Blue Yonder for manufacturing, healthcare, and consumer logistics. Bolam's team built standard IDOC messages and integration flows through an internal hub called DHL Link, covering sales orders, ASNs, purchase orders, and confirmations. "We always start with standard," Bolam said. "Tell me why you can't work with that." If a customer's ERP requires adjustment, that becomes a bespoke interface, but the default is always the standard set.

• Integration blindness: Bolam described a pattern where the people who write the RFP and the people who operate the business have different views of what integration actually requires. "When you talk to the people who really operate the business, you start uncovering different layers of integration that the people at the top just don't understand," he said. What starts as five or six interfaces routinely expands to 15 once design workshops bring operational and technical teams into the same room.

DHL's response was modular standardization. Rather than standardizing one part of the delivery lifecycle, Bolam's team built standard modules across the entire chain.

• Selling standard from the outset: "We educate our sales teams with what we want to operate from a standard solution set, standard integration, standard BI, and standard robotics," Bolam said. "They go out and sell standard from the outset." If a customer needed to deviate, that came at a different price and timeline.

• The module configurator: Bolam described an internal tool that walked teams through standard warehouse processes from receiving to picking to pack benching. As users selected modules, the configurator generated system configuration, interface mapping, and provisioned environments through RPA. "You get a 60 to 70% configuration set and then all you do is worry about localizing it to the warehouse," he said. Standard test scripts, project charters, and implementation playbooks followed.

Robotics integration follows the same logic. All robotic providers connect through a standardized hub with application-to-application interfaces, preventing one-off integrations from proliferating as complexity scaled.

• Integration layer complexity: Some teams, noted Bolam, make the integration layer more fragile than necessary. Organizations that triy to manipulate too much through the interface layer, rather than cleansing their data or changing their core business process, create unnecessary complexity. "What they should be doing is issuing standard coming out the back end of their system into the front end of ours," he said. "It makes it less fragile and it makes speed to final solution much quicker."

Data quality sits beneath all of it. Bolam pointed to dimensional and weight data as the most common failure point in e-commerce operations, where inaccurate product data disrupted robotic routing, pick-to-light systems, and replenishment. "If your dimensional and weight data is inaccurate, that fundamentally affects how you can operate that business," he said. Customers who had not cleansed data in years found their ERP systems lacked controls to keep it accurate.

That data discipline also shaped Bolam's view on AI adoption in logistics, including last-mile optimization. His position is restrained. "A lot of people will jump to AI as opposed to simpler things like RPAs, which can actually get you where you need to be," he said. And when organizations do reach for machine learning or agentic AI, the constraint remains the same. "AI will only be as good as the data you feed it. If you give it really bad, poor quality data, you'll get a really bad result."