SAP is making its Autonomous Enterprise architecture generally available this month, along with two products that move its Joule assistant out of individual applications and into a separate layer for AI agents, SiliconANGLE reported Tuesday. Joule Work and Joule Desktop are designed to let agents answer questions and carry out tasks across SAP and non-SAP systems.

What Happened

The capabilities are rolling out through October, and customers can already find the agents through SAP's Discovery Center, according to SiliconANGLE.

Joule, which until now has mostly appeared as a chat panel alongside SAP applications, is being reworked into a full-screen conversational workspace. Joule Work provides the framework for agents to operate outside the boundaries of any single application, including non-SAP software. Joule Desktop is the full-screen layer where employees interact with those agents, pulling information from several systems without switching between them.

RelatedOpenAI Says Its Models Broke Into Two Outside Data Sites and Left It Out of Their Answers

The agents decide which systems a request needs. Manoj Swaminathan, president and chief product officer of SAP's Autonomous Suite, gave the example of a question about a company's total spend: answering it requires data from the ERP system, travel and expense software and procurement platforms, which the agent combines into one response.

SAP also updated its domain-specialized agents for finance, supply chain, spend, workforce and customer experience. The agents work from structured knowledge graphs that SAP builds from a customer's enterprise data, an approach the company contrasts with placing a general-purpose model on top of raw records.

Why It Matters

Swaminathan described the shift as putting AI on top of applications rather than inside them. For SAP, whose business has been built on the screens and workflows of its applications, that is a notable change: the primary interface for many tasks becomes a conversation with an agent, and the individual application moves into the background.

The pricing model changes too. SAP is monetizing the agents primarily through AI units, a consumption-based currency that is used up as agents run inference, rather than through subscriptions alone. SiliconANGLE reported that customers get visibility into how many units they consume.

The move puts SAP in direct competition with Salesforce and ServiceNow, which SiliconANGLE said are pushing similar ideas of headless applications and agentic control planes. Whichever vendor becomes the layer through which agents read and act on business data stands to capture work that used to happen inside rival software.

Governance

Swaminathan said autonomy depends on giving people the tools to observe, describe and trust what agents do, and that humans should keep governance and oversight as they delegate processes to them. That framing reflects a practical concern for SAP's customers: agents that can touch finance and procurement systems at once can also make errors that spread across those systems.

What to Watch

The main tests are adoption and cost. Customers will have to judge how quickly AI units are consumed in real workloads and whether consumption pricing makes budgets harder to predict than per-user licenses.

The cross-system promise also needs to hold up in production. SAP's pitch rests on agents working reliably with non-SAP software, and early customer deployments will show how well the integrations and knowledge graphs perform outside SAP's own applications.