AI on SAP ECC without the cloud commitment
SAP said in May 2026 that its AI assistant Joule reaches ECC and on-premises customers who commit to moving the majority of their landscape to SAP Cloud ERP; SAP's chief strategy officer put it as half of the maintenance spend. You do not have to be pushed into that decision. AI over your SAP system is possible without the commitment: custom Joule skills through SAP BTP, or a Claude layer over your SAP data through MCP, with permissions scoped per deployment, personal data masked and a person approving every write.
Two paths, depending on your SAP
- Cloud editions (RISE, GROW): native Joule. Activation, configuration and custom extensions, including agents built in Joule Studio.
- ECC or on-premises: the third way, without a commitment. Custom Joule skills through SAP BTP, or a Claude layer over data from SAP through OData and RFC.
Processes where it pays from the first month
- Exceptions on incoming invoices: the agent compares invoice, purchase order and delivery note, prepares the posting and presents the exceptions for approval.
- Master data cleansing: duplicates, incomplete records and inconsistent fields found and prepared for correction, with a person confirming each change.
- Month-end close preparation: reconciliations, variance checks and management overviews, with links back to the source data.
The rule that does not change
The AI layer reads SAP through a scoped technical account with only the roles the use case needs, limited to data every pilot user may see; propagation of the individual's own SAP identity is on the roadmap and is stated as a limit in every proposal. Every write to the system is prepared by the layer and approved by a person before it runs. This is our standard, not a paid extra.
The integration runs through MCP, an open protocol through which the AI reads data and prepares actions in SAP and in the systems around it. Because the protocol is open, the layer is not tied to one vendor and can be moved when your SAP roadmap changes.
What we ask in the first conversation
Which SAP edition you run and what your maintenance roadmap looks like. Where you run cloud and company identities today. Whether you require data processing in the EU. Which two or three processes hurt most this quarter. From those answers we propose a path that a later change of decision does not turn into a new project.
Questions we get asked
Is the third way supported by SAP?
SAP BTP is SAP's own extension platform and custom skills on it are a supported pattern. A Claude layer over SAP data uses the OData and RFC interfaces SAP provides; the release status of each interface, the RFC implementation and the licensing position are confirmed with your Basis team and your SAP account team before production. Neither path modifies the ECC core.
Will this block a later move to S/4HANA?
No. The AI layer sits beside SAP and reads through standard interfaces. When you move to S/4HANA, the layer is re-pointed at the new system's interfaces and re-validated, and native Joule becomes an option alongside it.
Can the agent post documents in SAP?
It can prepare them. Every write is approved by a person, and the audit trail records the proposal, the approver and the change. Fully automatic posting is a decision you make per process, later, with the record to support it.
Does our data leave the company?
That depends on the deployment you choose. With Amazon Bedrock in Frankfurt or Google Vertex AI in EU regions the model runs under your own cloud account; retention and logging depend on the endpoint and are documented per component. Personal data is masked before it reaches the model in every variant.
Talk to us.
A 30-minute conversation on your scenarios. The call and the written assessment you leave with are free, whether we work together or not. The next step, if there is one, is a fixed-price six-week pilot in your environment with no obligation to continue.
Or write to us directly: info@iteractive.ai · +421 905 452 325