SAP has detailed its Q2 2026 Business AI release, moving several autonomous agents into general availability and attaching named customer results to them — an attempt to shift the enterprise-AI conversation from demos to measured outcomes.
Agents that run themselves
The release promotes multiple autonomous agents. A Project Billing Price Verification Agent (in beta) is said to cut resolution time by 75%, and an Order Reliability Agent (beta) to reduce analysis time by about 50%. Two reach general availability: an Expense Automation Agent, delivering up to 30% faster expense reports, and a Process Consulting Agent, cutting data-search time by up to 90%.
Joule gets hands and a voice
SAP's assistant, Joule, gains Joule Work (early adopter) plus a generally available mobile app, a Voice for Joule interface built with LiveKit, and Joule Studio for building custom agents. Together they push Joule from a chat helper toward a platform on which customers assemble their own automation.
The customer scorecard
SAP put numbers against real deployments: Bosch Digital reported a 20% developer productivity gain, Aeropuertos Argentina a 16% cost reduction, PwC handling VAT for a pharma client 60% more efficiently, and retailer LC Waikiki cutting HR cycle times by 40-60%. The figures are vendor-supplied and not independently audited — a caveat that applies to most enterprise-AI ROI claims — but naming customers raises the stakes on their accuracy.
The land SAP is defending
Written up by SAP Chief AI Officer Jonathan von Rueden, the release reflects a strategic reality: SAP sits inside the core systems — finance, procurement, HR — where enterprises run their operations. Embedding agents that act on those records directly is its bid to keep AI value flowing through its platform rather than through a separate orchestration layer bolted on by a rival.