SAP TechEd 2024: Joule Gets Collaborative Agents and a Knowledge Graph
October 2024: SAP moved Joule from a single copilot to collaborative agents that reason and act across finance, supply chain, and HR together, grounded by a new SAP Knowledge Graph.
SAP TechEd 2024: Joule Gets Collaborative Agents and a Knowledge Graph
IDC's Ritu Jyoti, on the shift: SAP's move "will enable agentic AI solutions that seamlessly integrate across the business to transform the future of work."
At SAP TechEd on October 9, 2024, SAP announced the step that turned Joule from a single question-answering copilot into a set of collaborative AI agents — the architecture everything SAP has shipped since traces back to.
The development
- Joule's agents were positioned to operate across business functions — supply chain, finance, HR — and coordinate with each other rather than each handling an isolated task.
- Headline example: a payment-dispute scenario where a collection agent sets the strategy, then hands off to specialized email, invoice, and support-case agents to execute the workflow end to end.
- Alongside the agents, SAP introduced the SAP Knowledge Graph — infrastructure meant to connect raw data with the business context captured inside SAP applications (e.g., linking a disputed payment to its related orders, invoices, and prior communications).
- Announced timeline: first Joule agents, plus extensibility tooling, available Q4 2024; more out-of-the-box agents and a "Joule studio" for customer-built agents promised for 2025.
What it actually means
- This is the announcement where "agent" replaced "copilot" as SAP's core AI framing — worth remembering when evaluating how long the current agent architecture has actually existed versus how new SAP's marketing makes it sound.
- A knowledge graph is the un-glamorous prerequisite for any of this working: agents reasoning over ERP data need structured business context (chart of accounts, order-to-cash relationships), not just raw table access. SAP naming this explicitly in 2024 is the origin of the "rich context layer" language used in SAP's 2026 materials.
- The collections example is deliberately chosen — payment disputes are a well-bounded, rules-legible process, which makes it a good demo but a poor proxy for messier processes like demand planning or workforce scheduling.
- "Extensibility capabilities" ships alongside the agents themselves, signaling early that SAP expected partners and customers to build on top rather than treat the initial agent set as complete.
What to watch for
Whether the "Joule studio" for customer-built agents, promised here for 2025, shipped as broadly as described — check it against later announcements before assuming custom agent-building is mature.
Related reading
For where the agent layer stands today, see our Joule & AI Agents hub. For plain-language breakdowns of each major SAP AI term, see SAP AI Explained.
Agents are only as good as the data context feeding them — data readiness is the first dimension our free AI Readiness Check scores in your own environment.