SAP Wraps Up Business AI Quarter: Agents Cut Process Costs by Up to 90 Percent

SAP published its quarterly Business AI report on July 20, 2026, bringing together for the first time hard numbers showing how much companies are actually saving by using AI agents built into the Joule system. The report covers releases from April to June 2026 and was signed by Jonathan von Rueden, Chief AI Officer at SAP SE.

What the Report Shows

SAP's quarterly roundup is not the announcement of a single product but a review of dozens of features rolled out over three months as part of the Business AI suite, covering the Autonomous Finance, Autonomous Spend, Autonomous CX and Autonomous SCM modules. At its center remains Joule, SAP's AI assistant, which gained a new iteration this quarter called Joule Work.

Joule Work is meant to change how employees carry out business processes, shifting the interface away from fragmented, transactional screens toward a single, unified work environment spanning both SAP systems and applications outside the ecosystem. A command expressed in natural language is meant to trigger teams of specialized agents coordinated by Joule assistants.

Voice and Phone, Not Keyboard

SAP also rolled out Joule Work as a mobile app now generally available, letting users handle sales, HR and supply chain processes from their phone, including via Siri voice commands on iPhone and iPad. In parallel, an early access program is underway for Voice for Joule, built in partnership with LiveKit, meant to let field workers talk to the system in real time instead of typing commands.

Another new feature is support for multiple SAP S/4HANA Cloud systems within a single Joule formation, allowing processes that span several separate installations to run without switching between interfaces.

Numbers From Customers

The most concrete part of the report is the results reported by SAP customers. Argentine airport operator Aeropuertos Argentina said the SNOW agent cut direct costs by 16 percent and reduced administrative workload by as much as 90 percent. Consulting firm PwC reported that its clients process VAT transfers 60 percent more efficiently thanks to AI agents built into finance processes.

Clothing retailer LC Waikiki reported HR process cycles shortened by 40 to 60 percent, while Bosch Digital reported a 20 percent productivity increase for development teams using Joule and 15-20 percent faster unit testing.

The report shows how SAP's AI solutions deliver measurable business value across different areas of our customers' operations - Jonathan von Rueden, Chief AI Officer, SAP SE

New Features Across Modules

In the Autonomous Finance module, a beta agent for verifying prices in project billing appeared, cutting the time needed to resolve pricing discrepancies by 75 percent. Autonomous Spend gained a generally available expense automation agent that cuts the time to complete expense reports by up to 30 percent, and Autonomous CX received an order reliability analysis agent that reduces the time spent analyzing unusual orders by about 50 percent and cuts customer churn by about 20 percent.

SAP also expanded the Business AI platform layer with Joule Studio, a new environment for building AI extensions available through an early access program, and with a batch API feature in the Generative AI Hub, designed to handle large volumes of queries to foundation models along with centralized logging and a feedback mechanism.

What It Means for Polish Companies

For Polish corporate customers using SAP systems, the report signals that agentic AI in ERP tools is no longer an experiment but is becoming part of the standard offering, measured in hard indicators of time and cost savings. Companies planning deployments can already check the availability of individual agents in their S/4HANA Cloud instances, since some features have reached general availability rather than remaining in pilot programs only.

The expanded support for external models, including Claude Opus 4.7 and Gemini 3.1 Flash Lite, also means greater flexibility for Joule customers in choosing a language model provider, rather than being tied to a single AI partner.