AI agents, into production on SAP.
SAP is putting AI agents into finance, supply chain, sales and service. Whether they help you depends on three things: which agents your edition actually includes, whether your data and integrations are clean enough for them to work, and who keeps them running after go-live. We handle all three, as a simple ladder: audit finds what's worth doing, build proves it in production, operations keeps it paying off.
As of November 2026.
Last verified 1 November 2026. Dates change; we re-check them before every engagement. All SAP deadlines →
Watch a governed agent handle an invoice exception.
This is how an accounts-payable agent behaves in practice. It reads the invoice, runs the three-way match against the purchase order and goods receipt, stops at anything outside tolerance, and hands the decision to a person before anything posts to SAP. The gate in the middle is the point: the agent does the checking, a human owns the call.
Illustrative demonstration with sample data. Not a live system or a client project.
Audit. Build. Operate.
AI & clean core readiness audit
Know what to build before you spend money building it. A short, fixed-fee assessment that answers three questions: is your data ready for agents, is your core clean enough to extend safely, and which workflow gives you the fastest, safest return.
Agent build: fixed-scope pilots
One workflow, built properly and put live. We take a single high-value workflow from your audit and build a governed agent, grounded in your SAP data, clean-core-safe, with monitoring and an audit trail from day one. Where SAP already delivers an agent for the job, we configure and connect it instead of building from scratch.
Agent operations: managed service
Who runs your agent after go-live? Traditional SAP application support watches ABAP and interfaces. It doesn't watch what actually breaks an agent: model changes, token-cost creep, wrong answers, grounding data going stale, and governance gaps.
Five checks we run.
- 1
Is the data ready?
Grounding sources complete, current and clean.
- 2
Will it survive the next upgrade?
Built on released APIs and side-by-side extensions.
- 3
Can you see what it does and costs?
Monitoring of quality, cost and drift, with an audit trail.
- 4
Is the governance real?
Clear boundaries, and a person approving the steps that matter.
Check five: who runs it after go-live? An owner for drift, cost, grounding refresh and audit. Read: Why most SAP AI agents never leave the sandbox →
Joule where it fits. Open standards where they don't.
We build on SAP Joule Studio where it's the right fit, and on open standards, Agent2Agent (A2A), Model Context Protocol (MCP) and a choice of language models, where they are. Your agent strategy isn't tied to a single product.
Integration Suite and BTP are the layer that lets agents reach your systems, data and each other. That's work we've done for 15 years. SAP Integration Suite →
Find out which SAP AI agents fit your landscape.
An AI readiness assessment maps SAP's agents to your edition, scores your candidate workflows and recommends a first pilot.