Our Take
SAP has built real infrastructure (a tabular data model, knowledge graph, and Joule Studio) but is using it to optimize what already exists rather than redesign how enterprise operations work.
Why it matters
Enterprise leaders betting on AI to fundamentally reshape operations need to understand the difference between making existing workflows faster and reimagining them entirely. SAP's approach delivers value now but leaves ROI on the table.
Do this week
SAP customers: audit your top 5 business pain points this week to determine which need automation versus architectural redesign, so you know whether to adopt SAP agents as-is or invest in custom builds via Joule Studio.
224 agents, but in service of existing workflows
SAP announced The Autonomous Enterprise, a new AI architecture led by CEO Christian Klein that adds a unified platform for building, contextualizing, and governing agents across the company's ERP suite. The announcement includes 224 agents and 51 assistants across HCM, finance, supply chain, and other modules.
The technical foundation includes three significant new components. SAP-RPT-1.5, a proprietary model optimized for tabular data analysis, allows AI to work with the spreadsheets and tables that form the backbone of enterprise software. A semantic layer called SAP Knowledge Graph maps business entities, rules, and structures so natural language queries can be translated into function-specific searches. Joule, originally a chatbot interface, is now a full agent development environment that allows IT teams and developers to design, test, and deploy agents across the SAP ecosystem.
In practice, the agents focus on automation: finding payment errors in payroll processing, generating employee development content, identifying supply chain cost drivers, and flagging operational anomalies. A profit-margin question that once required human analysts can now be answered by Joule in seconds by tracing supplier price changes, shipping costs, and commodity inputs across the entire business.
Optimization without transformation
The distinction SAP is making—and the one practitioners often miss—is between two modes of AI deployment. The first automates existing processes: make the same workflow faster, cheaper, cheaper, or error-free. The second redesigns the workflow itself based on what becomes possible with AI.
SAP is currently in stage one. Its agents take the current ERP architecture as fixed and find ways to operate it more efficiently. This is valuable. Complex ERP systems are notoriously difficult to navigate; automation that surfaces hidden cost drivers or fixes payment errors without manual intervention adds real money to the bottom line.
But SAP's Bersin's research model shows stages 3 and 4 (strategic redesign and continuous optimization) deliver 5 to 10 times higher ROI than automation alone. SAP has built the technical scaffolding to reach those stages. The knowledge graph and large context window needed to load entire company memory already exist. Joule Studio is advanced enough to allow custom agent development. But the announced use cases and demos remain focused on making the existing machine run better, not reimagining the machine itself.
For a company like SAP—which manages financial transactions, human capital processes, supply chains, and vendor relationships across industries as varied as pharma and airlines—the gap is substantial. A pharma company could use the same data layer not to automate its current hiring process but to build an entirely new talent-matching system based on skills, tenure, market dynamics, and organizational structure. Instead, the emphasis remains on agent-driven monitoring and error correction.
Where to invest your attention
SAP customers should examine the 224 agents and identify which address genuine pain points versus which automate manual workarounds that exist only because the software is hard to use. WalkMe, a process automation tool that SAP already owns, can handle many of these workflows without a new agent architecture.
The real strategic opportunity lies in Joule Studio and the knowledge graph. If your organization has unique business logic or process requirements that the standard agents don't cover, Joule Studio offers the tools to build custom agents. This is where process redesign becomes possible. The technical debt is high—ERP systems are notoriously rigid—but the upside of redesign is far larger than optimization of the status quo.
Watch Workday's Sana Studio and ServiceNow's agent announcements over the next 12 months. Both have made similar architectural moves and will face the same choice: continue automating legacy processes or invest in redesign workflows. Whichever vendor moves first into stage 3 and 4 use cases, with published customer results, will set the pace for the industry.