SAP Brings Autonomous Enterprise Out of Apps With Expanded Joule Platform

SAP SE is moving beyond autonomous AI assistants confined to individual applications, instead creating a framework for agents to operate across an organization's entire software ecosystem. At SAP Connect 2026, the company made its Autonomous Enterprise architecture generally available this month, rolling out Joule Work and Desktop alongside an expanded collection of autonomous assistants and agents.

Manoj Swaminathan, SAP's president and chief product officer of Autonomous Suite, explained that autonomy represents a gradual transformation rather than an overnight shift. "When we talk about autonomous enterprise, we realize that it's going to be a journey for the customer, and it's not something you draw a line to say starting tomorrow will be autonomous just because you now have these agents and assistants," he told SiliconANGLE.

Joule, SAP's generative AI assistant launched in 2023, has evolved into the company's primary autonomous tool. The latest enhancements position it as a conversational interface layer that operates alongside applications rather than embedded within them, functioning as both an automation and productivity engine.

Swaminathan emphasized that true autonomy requires human oversight and control. Rather than simply delegating all tasks to agents, the approach gives users the ability to observe, describe and establish trust in agent behavior before handing over processes. This includes the capacity to manage how agents interact with data, applications and enterprise context, plus the transparency needed to monitor and review their actions.

First principle is it's AI on apps as opposed to AI in apps

Manoj Swaminathan

Joule Work and Desktop: Operating Beyond Application Boundaries

Joule Work establishes the infrastructure for deploying AI assistants into customer workflows, enabling autonomous capabilities to function with both SAP and third-party applications. While existing AI agents operate within individual apps—such as drafting documents in word processors or retrieving email history—SAP's new layer connects across enterprise systems without requiring users to switch between them.

Swaminathan illustrated this capability with a corporate spend example. Rather than treating the question as isolated to one system, Joule can recognize that answering it requires pulling data from ERP, travel and expense systems and procurement platforms simultaneously. The agent then consolidates these records and delivers a unified business answer.

Joule Desktop serves as the interface layer where customers spend their time, extracting the same enterprise information from their applications. It functions as an intelligent overlay capable of handling work-related tasks without tying them to individual application resources.

Joule is no longer a natural language chat client for us; it is a full-screen experience

Manoj Swaminathan

The general availability launch includes refreshed agents specializing in finance, supply chain, spend management, workforce and customer experience, along with other industry-focused variants. These agents bring domain-specific expertise, knowledge and business judgment to enterprise data interactions.

Moving from AI Models to Autonomous Enterprise

Swaminathan noted that enterprise customers typically supply their own data and cloud data lakes, while SAP structures that information and delivers it through a knowledge graph. Off-the-shelf models from OpenAI Group PBC and Anthropic PBC lack the ability to provide this depth when simply layered onto enterprise data. ERP semantics, processes, policies and relationships represent what Swaminathan called the "secret sauce" that differentiates SAP's approach.

SAP is not alone in this shift. Salesforce Inc. and ServiceNow Inc. are also advancing toward headless applications, context layers and agentic control planes. The industry is moving away from adding copilots to applications themselves and toward constructing an intelligent layer that sits above the customer experience.

Swaminathan stressed that SAP customers gain the ability to express intent while agents determine which systems to involve and execute the work. Humans remain central to the process, maintaining governance, auditability and visibility, with agents functioning as employees within the system—subject to the same security, systems of record and execution standards. Agents represent an additional intelligent conversational layer operating on behalf of humans rather than replacing them.

Autonomous enterprise is not a vision anymore. It is real, with general availability in October.

Manoj Swaminathan

Monetization Through AI Units

SAP will continue leveraging AI units as its primary commercial model for these capabilities. The approach emphasizes consumption-based pricing rather than pure subscription models, with agent inference drawing from the unit pool. SAP will give customers visibility into their usage patterns.

Joule updates and other autonomous enterprise features are rolling out throughout October. Customers can already explore agents through SAP's Discovery Center.