SAP and Intel are testing a system that turns warehouse tasks created in SAP into missions carried out autonomously by robots at Intel’s Chengdu site in China. It is a concrete step in physical AI: connecting business software to actions in the real world, rather than leaving AI to make recommendations on a screen.
SAP Watch analysis
What happened
In the proof of concept, SAP provides the business workflows and context, while Intel supplies edge computing and robotics technology. Warehouse transport tasks generated in SAP are translated into robot missions and executed in the warehouse. Diginomica reports that the collaboration was presented at SAP TechEd; the site disclosed that SAP was a premier partner at the time of publication.
SAP describes its wider physical-AI approach as business-led, traceable and designed to work across different hardware and AI models. Its executives say customers are already piloting the capabilities, but the Chengdu example remains a proof of concept, not evidence of a scaled commercial deployment.
Why it matters
The practical ambition is to connect operational software with physical work: a warehouse task in SAP could become a robot’s instruction, with activity recorded back into business workflows. If that works reliably beyond a demonstration, it could help companies coordinate routine material movement and reduce repetitive manual handling.
SAP executive Manoj Swaminathan says unplanned downtime costs the company’s largest customers 11% of revenue. That is SAP’s account of the problem, not an independently established measure of what this project will save. The report does not give a quantified result for the Chengdu proof of concept.
Our read
This is a more persuasive physical-AI pitch than a robot wandering around a stage while someone calls it a strategy. The interesting part is the connection between enterprise instructions and robot actions, with SAP retaining the workflow context and Intel providing the edge technology.
The proof now has to leave the demo lane. A useful test is whether the system can handle real warehouse variation safely and reliably, and whether customers can measure gains in throughput, downtime or labour. Until those results arrive, this is a promising integration, not a finished business case.
What to watch
- Whether SAP and Intel report measured results from the Chengdu proof of concept.
- Whether the system moves from demonstration to routine warehouse operations.
- How the companies handle exceptions, safety and human oversight.
- Whether other hardware and robotics platforms can use SAP’s approach.
Discussion spark: Should enterprise software vendors control the business rules behind warehouse robots, or should customers require those rules to stay independent of the vendor?
Sources and evidence
- Want to hear more about Enterprise AI? (28 October 2026, 09:05 UTC)
not affiliated with or endorsed by SAP