Run, support, and evolve your SAP landscape with governed AI, from insight to change. Talk to SAP in everyday language, with SAP-grade authorisation, governance and auditability on every action.
Keep scrolling.
Short dumps with source extracts and call stacks, gateway error logs, system messages, failed IDocs, batch jobs and the security audit log, read through SAP-native interfaces that work under standard authorisations, then explained and correlated in plain language by the AI.
Read tables, run sums, counts and group-by aggregations pushed down into SAP, and call OData services and RFC function modules, with results returned as plain-language answers that flag what looks unusual. Read-only roles are enforced by a policy engine at the boundary, on every access path.
Proposed changes are validated with syntax and ABAP Test Cockpit checks before any write. Applies run a full lock, edit, activate cycle with automatic recovery on activation failure. Transport release and deletions require an explicit human action the AI cannot invoke, and every step lands in the audit trail.
Scaffold the full RAP stack from a classic Z-table or report, with CDS views, behaviour definition and class, and service definition and binding, publish a live OData V4 service, then generate and deploy a Fiori Elements app hardened with SAP’s official UI5 tooling. Weeks of boilerplate become a governed, reviewable starting point.
Every interaction, read or write, lands in a complete, append-only audit record: the real user’s identity, the parameters, the SAP response, and the decision of every governance gate.
Ask me anything: explain an object, find custom code for a transaction, analyse dependencies, or propose a change.
What went wrong in S4D in the last 24 hours? Short dumps, failed IDocs, cancelled jobs, worst first.
08:02
No business-critical failures overnight. Worst first: a HANA connection drop during a monitoring task (transient, one for Basis only if it repeats), a flood of RFC authorisation dumps from a technical user, and one batch job aborting daily since March. IDoc processing is fully healthy.
Category ABAP programming errorRuntime error DBSQL_SQL_ERROR · SQL code -10807Connection down: 'send' failed, rc=104: Connection reset by peerProgram SAPLTASK_VITAL · User SAP_WSRT| PRIO | FINDING | OWNER |
|---|---|---|
| P1 | HANA connection reset mid-query (transient) | Basis |
| P2 | SM_SMP missing RFC auth, dumping every ~5 min | Security |
| P3 | Daily job aborting at 09:00 since March | BWALKER |
| OK | IDocs: no failures in 24 h | n/a |
Fastest win: SM_SMP has no roles assigned at all. Assign the existing ZSAP_SOLMAN_READ_702 role in client 110 and the dump flood (about 300 a day) stops.
0 in · 0 out · 0 cached · 0% cache hit
Count documents by type and posting period for company code 2300, fiscal 2026, and flag anything unusual.
08:03
Live aggregate from S4D, counted inside SAP with only the totals returned. One thing stands out: 91% of FY 2026 volume sits in period 0 with a blank document type. That is a migration carry-forward load, not real activity. True transactional volume is about 127k lines.
| PERIOD | DOC LINES | SHARE | FLAG |
|---|---|---|---|
| 0 · Carry-fwd | 1,284,516 | 91.0% | Migration load |
| 1 · Jan | 19,204 | 1.4% | |
| 2 · Feb | 2,871 | 0.2% | Low |
| 3 · Mar | 81,340 | 5.8% | Spike |
| 4 · Apr | 5,512 | 0.4% | |
| 5 · May | 703 | 0.05% | Very low |
| 6 · Jun | 2,988 | 0.2% |
0 in · 0 out · 0 cached · 0% cache hit
The partner lookup in ZTRADESYNC never checks sy-subrc, so a missing business partner slips through. Propose a fix, validate it, and apply it.
08:04
The result of the partner SELECT SINGLE is used unchecked, so a missing business partner proceeds with stale data. The guard below returns a clear error instead. Syntax, ATC and abaplint checks passed. Review the diff and approve to apply it on S4D.
Change
Guard the business-partner lookup: return a clear error when the partner does not exist instead of proceeding with stale fields.
Diff (12 lines)
Transport Request
S4DK903117 · Trading-platform BP sync fixesChange
Guard the business-partner lookup: return a clear error when the partner does not exist instead of proceeding with stale fields.
Diff (12 lines)
Trading-platform BP sync fixes
Releasing this transport moves its objects out of DEV toward test/prod and cannot be undone. This is the human signoff, an explicit action the AI cannot invoke.
Your customer-namespace ABAP objects, grouped by type. Pick an object and the assistant will analyse it end-to-end: classic-API dependencies, clean-core status, and a suggested path to the modern SAP stack (RAP / CDS / Fiori).
Analysis only. No source is modified until you approve a specific change in chat.
| Name | Description |
|---|---|
| ZTRADESYNC | Trading-platform BP integration (SEGW core) |
| ZCL_ZAPI_TSY_DPC_EXT | Gateway data provider extension |
| ZCL_ZAPI_TSY_MPC_EXT | Gateway model provider extension |
Every governed action (apply, reject, activation, ATC outcome) recorded with timestamp, request ID, and policy decision.
| Timestamp | Tool | Status |
|---|---|---|
| 2026-08-24 08:02:14 | list_short_dumps | success |
| 2026-08-24 08:03:22 | query_aggregate | success |
| 2026-08-24 08:04:36 | validate_change_set | success |
| 2026-08-24 08:04:52 | apply_patch_with_activation | success |
| 2026-08-24 08:05:10 | release_transport | error |
{
"transport": "S4DK903117",
"requested_by": "agent"
}Short dumps with source extracts and call stacks, gateway error logs, system messages, failed IDocs, batch jobs and the security audit log, read through SAP-native interfaces that work under standard authorisations, then explained and correlated in plain language by the AI.
Read tables, run sums, counts and group-by aggregations pushed down into SAP, and call OData services and RFC function modules, with results returned as plain-language answers that flag what looks unusual. Read-only roles are enforced by a policy engine at the boundary, on every access path.
Proposed changes are validated with syntax and ABAP Test Cockpit checks before any write. Applies run a full lock, edit, activate cycle with automatic recovery on activation failure. Transport release and deletions require an explicit human action the AI cannot invoke, and every step lands in the audit trail.
Scaffold the full RAP stack from a classic Z-table or report, with CDS views, behaviour definition and class, and service definition and binding, publish a live OData V4 service, then generate and deploy a Fiori Elements app hardened with SAP’s official UI5 tooling. Weeks of boilerplate become a governed, reviewable starting point.
Ask why an invoice failed to post, trace an IDoc through the chain, follow an integration message end-to-end, answer a business question from live SAP data, or modernise a legacy Z object, all through one audited platform. People receive clear, human-readable answers covering exactly what their own SAP user is authorised for, and nothing beyond it. Governance is enforced in code, not in prompts.
Diagnose live issues in minutes: short dumps with source extracts and call stacks, gateway errors, failed IDocs, stuck batch jobs, system messages and integration message logs, all explained and correlated in plain language. No SAPGUI archaeology, no waiting on a developer.
Safe, self-service answers from live SAP data: did an invoice post, where is a purchase order, what are this year’s totals by account. Aggregations run inside SAP, and a role-based policy engine keeps these users strictly read-only, so self-service never becomes uncontrolled access.
Understand, validate, change and modernise the custom estate: from explaining undocumented Z code and tracing dependencies, to governed ABAP changes, to scaffolding a classic Z-report into a modern RAP application with a live OData service and a generated Fiori app.
One audited tool surface for diagnostics, live data, integration and engineering, available from the platform's own chat UI and, via the open Model Context Protocol, from AI tools your team already uses. Identical governance and identical audit entries on every path.
Every call is scoped to the individual user’s own SAP authorisations, and the controls are enforced in code, not in prompts. The AI never decides what a user may see or do.
Every proposed change is validated with syntax and ABAP Test Cockpit checks before anything is written. Releasing a transport towards test and production, or deleting an object, remains an explicit human action the AI cannot invoke.
Every interaction, read or write, lands in a complete, append-only audit record: the real user’s identity, the parameters, the SAP response, and the decision of every governance gate.
Customer-hosted on SAP BTP Cloud Foundry, in your own subaccount. SAP data never transits ETZ infrastructure. You bring your own AI provider, so model choice and data residency stay in your hands.
FAQ Questions
If your question isn't covered here, reach out and our team will get back to you.
Start a ConversationA governed AI runtime for your SAP landscape. People talk to their SAP system in everyday language and receive clear, human-readable answers, covering exactly the information and capabilities their own SAP user is authorised for, and nothing beyond it. It serves the whole SAP team: operations and support, functional consultants, business users, and developers.
The two are complementary. Joule assists business users inside SAP applications; the ETZ AI Platform is a governed AI runtime for the teams that run, support, build and modernise the SAP landscape itself.
The platform is customer-hosted on SAP BTP Cloud Foundry. SAP data never transits ETZ infrastructure. The AI provider is your own account (Anthropic, OpenAI, or Azure OpenAI), so model choice and data residency remain yours. Deployment on Kubernetes or Docker is available for demonstrations and proofs of concept, and air-gapped deployment can be scoped as a bespoke arrangement with ETZ.
Diagnose incidents across the SAP stack, answer questions from live SAP data with in-system aggregation, operate SAP Cloud Integration end-to-end, explain and validate custom ABAP, execute governed changes, and scaffold classic Z objects into the full RAP stack with a live OData service and a generated Fiori Elements app.
Live-verified against SAP S/4HANA on-premise and private cloud, SAP BTP Cloud Foundry, and SAP Integration Suite (Cloud Integration). The platform talks to SAP through standard ADT, OData and RFC interfaces. Ask us about your specific landscape.
Yes, and the platform is built on that assumption. AI-generated answers and code proposals must be reviewed by a qualified human before being relied on. That is why every change is validated against SAP’s own checks before it can be written, why writes only ever reach development systems, and why promotion to test and production remains an explicit human decision.
We have a strong global presence, marking our commitment to delivering tech excellence in key markets worldwide.