SAP says its supply-chain AI agents are moving beyond recommendations towards carrying out work inside business systems, and points to customer deployments as evidence of progress. One customer, Lemvigh-Müller, has three coordinated agents handling supplier order confirmations, with SAP reporting more than 90 per cent touchless processing of 100,000 confirmations a year.
SAP Watch analysis
What happened
In a 3 November post, SAP said Gartner had placed it in the Market Shapers quadrant of its inaugural assessment of integrated supply-chain AI agents. The post describes Joule assistants coordinating agents across planning, sourcing, manufacturing and delivery, with agents acting within business-defined guardrails.
SAP also describes Takeda using agents to monitor manufacturing performance and start corrective workflows, with expected productivity gains of up to 10 per cent. RWE is applying agents to offshore wind maintenance, where SAP says they analyse past incidents and prepare work orders. These examples and figures come from SAP’s account.
Why it matters
Supply chains connect decisions across suppliers, factories, logistics and customer commitments. Software that can act across those processes, rather than simply flag a problem, could reduce routine handling and help teams respond to disruptions. The Lemvigh-Müller figure offers a concrete measure of one such deployment; the Takeda figures are expectations, not reported results.
Our read
The customer examples are more useful than the Gartner badge: they give readers something specific to judge, even if the figures come from SAP. The key question is whether agents can take on routine work without making oversight harder when a decision crosses several business systems. “Autonomous” sounds impressive; dependable performance is the bit that earns its keep.
What to watch
- Whether Takeda reports results against the expected productivity and stock-out figures.
- Whether Lemvigh-Müller’s reported processing rate holds as the deployment scales.
- How SAP explains the permissions, audit trails and human oversight for agents acting in live workflows.
Discussion spark: When supply-chain agents can execute tasks across business systems, which should matter more: reducing routine human workload, or requiring human approval at each step?
Sources and evidence
not affiliated with or endorsed by SAP