Intel and SAP Put SAP EWM in Charge of Autonomous Robots: A Chengdu Proof of Concept, Explained

Warehouse automation rarely stalls for want of a robot. The harder problem sits above the machine floor: getting an enterprise system to issue the task, and getting the machine's answer back into the system of record. A joint demonstration by Intel and SAP addresses precisely that seam. Shown during SAP TechEd and run on Intel's own premises in Chengdu, China, the proof of concept ties SAP Extended Warehouse Management — the platform most teams simply call EWM — to autonomous robots. Transport requests travel one way, task status and completion records travel the other, and the round trip closes inside SAP EWM. The two companies frame the result cautiously, citing shorter transport distances and less repetitive manual handling, while publishing no interface specification. What follows is a breakdown of what has been stated, who supplies which layer, and what an operations or data team still cannot learn from the announcements.

The confirmed facts

Element	Publicly stated detail
Parties	Intel and SAP; SAP additionally names Anu Intelligence as a Physical AI ecosystem partner involved in the project.
Venue and site	Presented during SAP TechEd; the demonstration itself ran inside Intel's own facility in Chengdu, China.
Systems in the loop	SAP Extended Warehouse Management (EWM) on the business side; autonomous robots performing material transport on the operational side.
Signal direction	Outbound: warehouse transport requests converted into robot missions. Inbound: task status, completion, outcomes and exceptions reported back into enterprise systems.
Claimed gains	Reduced transport distances and reduced repetitive manual handling, according to the companies.
Measurements	The source statements describe the reduction without publishing supporting figures or baselines.
Undisclosed	Interfaces and messaging protocols, the AI models behind the behaviour, and the recovery path for failed tasks.
Source image credit	Intel Corp.

Step by step: the loop that was demonstrated

1. A warehouse transport request originates in SAP EWM.
2. That request is dispatched to an autonomous robot as a mission.
3. The robot physically moves materials between locations.
4. Task status and completion are reported back to the warehouse-management system.
5. Outcomes and exceptions likewise flow back into enterprise systems, in Intel's description.

For this particular demonstration the companies use a deliberately narrow description: autonomous material transport plus feedback into SAP EWM. Broader warehouse activities are documented elsewhere in SAP's own material rather than in this demo.

Who supplies what

• SAP: the business context, workflow governance and process rules that define what may be executed and under which conditions.
• Intel: edge computing plus robotics hardware and software used for the autonomous material transport itself.
• Anu Intelligence: identified by SAP as a Physical AI ecosystem partner involved in the project.

From the companies themselves

"Organizations want to connect AI directly to physical operations. By combining SAP's enterprise context and orchestration with Intel's edge AI and robotics technologies, we can help customers move Physical AI from business decision to real-world execution." said John Healy, Vice President and General Manager, Industrial and Robotics at Intel.

"Physical AI can reshape operations by connecting enterprise systems with operational technology. SAP and Intel are creating solutions to help customers integrate AI-driven automation into core processes while preserving flexibility across robotics platforms, deployment models and ecosystem partners across both hardware and software." said Lukasz Ostrowski, Head of Physical AI and Robotics, SAP Research and Innovation for SAP SE.

Design intent: openness instead of lock-in

• Architecture: characterised by Intel as open and robot-agnostic.
• Customer choice preserved across: robot vendors, operating systems, AI models, and cloud or on-premises deployments.
• Stated next step: evolving the proof of concept into a repeatable approach that can be deployed across sites and robot platforms.

The toolchain named alongside the effort spans OpenVINO Physical AI, Physical AI Studio and Robotics AI Suite. Intel's public Robotics AI Suite repository is described as providing applications, libraries, examples and benchmarking tools covering autonomous mobile robots, stationary arms and humanoid systems, with ROS 2 integration and OpenVINO-optimized AI models.

SAP publishes a separate warehouse Physical AI reference architecture that describes a wider set of operations, including picking, placing, sorting and packing — a scope broader than the transport-and-feedback loop shown in Chengdu.

Why enterprise readers should pay attention

The practical significance is the connection between warehouse-management decisions and autonomous execution, with task outcomes reported back into the system governing the work. Read that way, the demo is less about the robot and more about the audit trail: an enterprise system that both authorises a physical action and receives the record of whether it happened. For organisations building data governance around automation, that return path is the part that determines whether machine activity becomes queryable operational data or remains trapped on the plant floor.

What the announcements leave open

• The interfaces or messaging protocols used in the proof of concept are not specified.
• The AI models behind the demonstration are not identified.
• The mechanism for recovering or escalating failed tasks is not explained; Intel says exceptions return to enterprise systems but does not detail the recovery process.

Three questions a manufacturing IT team would ask next

1. How much integration work would another deployment require?
2. Which components can be reused across sites?
3. Do the reported improvements hold up under production conditions rather than demonstration conditions?

FAQ: common questions from data and operations readers

What exactly was connected?

SAP Extended Warehouse Management and autonomous robots. Transport requests issued by SAP EWM become robot missions, and status plus completion updates are returned to SAP EWM.

Is this a product launch?

No. Both companies describe it as a proof of concept, presented during SAP TechEd and demonstrated on Intel's own site in Chengdu, China.

Why does the return path into SAP EWM matter?

Because it closes the loop between planning and doing. Without it, the warehouse system issues instructions it cannot verify; with it, task status, outcomes and exceptions re-enter the system that governs the work.

Does the demonstration commit a buyer to a single robot vendor?

The stated intent is the opposite. Intel describes an open, robot-agnostic architecture meant to preserve choice across robot vendors, operating systems, AI models and deployment models. No specific robot fleet for the demo has been named.

Which robot vendors are supported?

The announcements do not name the vendors used in the proof of concept, though SAP identifies Anu Intelligence as a Physical AI ecosystem partner involved in the project.

What evidence of benefit has been published?

A qualitative claim only: the companies state that the proof of concept reduced transport distances and repetitive manual handling. No figures, protocol definitions or model names accompany that claim.

What to watch next

For now this is a plumbing story rather than a product story — an enterprise warehouse system learning to task robots and to listen for their reply, proven on its host's own floor before anywhere else. The signals worth tracking from here are a published interface and protocol description, a documented exception-handling and recovery path, and a second deployment outside the original site that could confirm whether the reported reductions survive production conditions. Until those appear, the sensible reading for decision-makers is architectural rather than commercial: the pattern of decision-to-robot-to-record is the takeaway, not a measurable return.

Compiled by David Nakamura for AI & Data News (AIdataNewsGlobal). Corrections or document leads: editorial@aidatanews.com.