Bounded autonomy
Agentic activity runs up to reviewable handoffs. It contributes across the engineering lifecycle but does not move an idea to a SAP write on its own.
KNOWLEDGE / AGENTIC ENGINEERING FOR SAP
Agentic engineering for SAP means AI-driven engineering activity that operates with bounded autonomy inside an explicit authority model — not open-ended automation of SAP changes.
Definition
In CAPELLA AFEP, an agent contributes across planning, generation and qualification, but SFC still qualifies, Govern still authorizes, A4Layer still executes and SAP still decides. Each stage has a named authority and a visible boundary, so activity never silently becomes an unsupported capability claim.
Agentic activity runs up to reviewable handoffs. It contributes across the engineering lifecycle but does not move an idea to a SAP write on its own.
SFC 0.13.1 compiles and qualifies managed RAP delivery candidates before any SAP-write authorization can exist.
Govern applies a hash-bound human approval bound to the exact qualified execution request.
A4Layer 1.0.0 carries approved intent into bounded execution, and SAP determines authorization, realized state, activation and runtime truth.
CAPELLA relationship
CAPELLA AFEP is the governed product surface for this pattern. Its one bounded verified outcome is a 14-object managed RAP realization that moved through SFC 0.13.1 qualification, explicit human authorization, A4Layer 1.0.0 execution, SAP readback and ATC with 0 findings. It does not claim unrestricted RAP delivery.
CAPELLA is an independent engineering project. It is not an SAP product and is not affiliated with, sponsored by, certified by or endorsed by SAP SE or its affiliates. “For SAP” describes the intended SAP software environment only.
See it in the product
Keep exploring
Architecture & authority model Verified evidence Ask CAPELLA