SAP SE announced a shift from promising agentic artificial intelligence autonomy to defining a path for users to interact with agents for work outside applications. During SAP Connect 2026, the company stated it is making its Autonomous Enterprise architecture generally available this month. The architecture was first introduced at Sapphire in May. It arrives alongside Joule Work and Desktop, featuring an expanded suite of autonomous assistants and agents.
Autonomous enterprise requires a customer journey
Manoj Swaminathan, SAP president and chief product officer of Autonomous Suite, told SiliconANGLE that achieving an autonomous enterprise is a journey. He noted that companies cannot simply draw a line and become autonomous overnight just because agents and assistants are now available. Autonomous operation does not mean handing everything to agents without supervision. Instead, the approach provides humans with tools to observe, describe, and trust agents before delegating business processes to them. This involves controlling, governing, and understanding how agents interact with company data, applications, and enterprise context before setting them loose, supported by transparency to audit their activity.
Joule operates across multiple applications
Joule debuted in 2023 as a generative AI assistant and now functions as an autonomous workhorse and conversational interface layer that works alongside apps. Swaminathan explained that the core principle involves AI operating on applications rather than strictly inside them. Joule Work provides a framework to bring AI assistants to customer workflows, spanning both SAP and non-SAP applications. For instance, answering a question about total corporate spend might require data from an enterprise resource planning system, travel and expense software, and procurement platforms. Joule can combine these records and present a single business answer.
Enterprise data provides the foundational context
Joule Desktop lives where customers interact with their applications, acting as an intelligent layer that performs duties directly related to the work. Swaminathan noted that Joule functions as a full-screen experience rather than just a natural language chat client. The October general availability includes updated agents for finance, supply chain, spend, workforce and customer experience, and other industry specializations. Swaminathan stated that enterprise customers bring their data and cloud lakehouses, while SAP organizes that data into a structured knowledge graph. General-purpose models from OpenAI and Anthropic cannot provide this capability when simply layered on top of enterprise data because ERP processes, semantics, policies, and relationships matter.
Salesforce and ServiceNow are also driving an industry trend toward headless applications, context layers, and agentic control planes. Enterprise AI is moving away from simply adding copilots into applications and toward building intelligent layers on top of the customer experience. Swaminathan emphasized that customers declare their intent, and the agent decides which systems to involve to finish the work. Humans remain in the loop with governance, auditability, and transparency, treating agents like employees equipped with security and systems of record. Monetization will rely primarily on consumption-based AI units rather than purely subscription-based pricing, with agent inference consuming units while SAP provides usage visibility.
Customers can already view the new agents in SAP’s Discovery Center, with the Joule updates and other autonomous enterprise capabilities rolling out throughout October.
