On October 6, 2026, at SAP Connect in Las Vegas, SAP said Joule Work is "beginning its customer rollout" as a new AI user interface, and that the Autonomous Enterprise architecture introduced at Sapphire in May becomes generally available during October. SAP's product chief for the Autonomous Suite, Manoj Swaminathan, summed up the design to SiliconANGLE: "First principle is it's AI on apps as opposed to AI in apps."
An enterprise example: what the workflow looks like
- One question, several systems. Swaminathan's example is total corporate spend: instead of treating it as one application's question, Joule works out that it needs records from ERP, travel and expense, and procurement, combines them and returns one answer. SAP says Joule Work is grounded in its SAP Knowledge Graph, which "maps more than 7 million data fields," and works across SAP and third-party data.
- From answers to actions. "Joule Assistants" carry out operational work "using established data, authorizations and audit trails," with customers controlling the level of autonomy. Through the Agent2Agent protocol, Joule can connect to third-party agents.
- Named customer uses (company-supplied quotes): Nestlé uses an Accounts Receivable Assistant to prioritise collections by risk; Novartis is piloting a Joule Sourcing Assistant it co-developed; PwC uses a Billing Assistant to catch billing errors earlier; Maschinenfabrik Reinhausen says lookups that "once required up to an hour of manual effort" across S/4HANA now take seconds.
- SAP as its own first customer. SAP says Joule Work is live with 110,000 employees internally and reports 20% productivity gains across finance, HR and procurement -- SAP's own figure, with no stated method.
The adoption constraints SAP is selling against
Governance. Swaminathan told SiliconANGLE that "autonomous" means giving people tools to observe, describe and trust agents before delegating, and that agents are treated like employees in the system, including security and systems of record. SAP points to an AI Agent Hub in SAP LeanIX as a central agent inventory and an ISO/IEC 42001-certified governance architecture.
Cost. SiliconANGLE reports that SAP will keep "AI units" as the main way it charges, "primarily consumption-based rather than purely subscription-based," with agent inference consuming units and SAP showing customers their usage. For a finance team, that turns an always-available assistant into a metered line item that grows with adoption.
Change management. Even a willing customer described work to get ready: CONA Services thanked SAP's team for "rapid remediation of defects" that let it become "Joule ready much sooner." Swaminathan himself framed it as "a journey for the customer."
Enterprise pattern (analysis): SAP, Salesforce and ServiceNow are all converging on a layer above the applications that knows the business's data model. SAP's case is that general-purpose models from OpenAI or Anthropic can't supply ERP semantics on their own. Buyers should test that claim on their own cross-system questions, and should model AI-unit consumption at full rollout before signing.
SAP began rolling out Joule Work on October 6 as an 'AI on apps' layer that answers and acts across SAP and third-party systems, claims 110,000 internal users and 20% productivity gains, and charges by consumed AI units -- making governance controls and metered cost, not model quality, the main things for buyers to test.