SAP Enterprise AI Needs Clear Responsibility
Agents, extensions, workflows, and core systems should not collapse into one blurred execution layer.
I help SAP teams reason about what belongs where:
- what should become an agent,
- what should remain a governed extension,
- what belongs in workflow,
- what must stay anchored in S/4HANA,
- and who owns truth, lifecycle, and accountability.
The Boundary Model™ originated from SAP BTP builder-side experience with post-go-live extension failures. In the AI era, it extends to agent-extension responsibility architecture.
Jiandong Pei is an independent SAP BTP & Enterprise AI architect, former SAP engineer, and creator of the Boundary Model™.
Former SAP engineer with builder-side exposure to SAP BTP platform and extension architecture, including CAP, XSUAA, AppRouter, Work Zone, multitenancy, integration, lifecycle, and SAP Build Code / Joule-era platform work.
The problem
AI can help teams create agents, workflows, integrations, and extensions faster than before. But faster creation does not automatically create clear responsibility.
A demo may work while truth, execution, lifecycle, and accountability are already drifting.
What I focus on
Agent–Extension Responsibility
Clarifying what belongs to agents, governed tools, extension APIs, workflows, assistants, and core systems before execution authority spreads informally.
SAP BTP Boundary Model
Runtime, identity, tenant, data, integration, and lifecycle boundaries in SAP BTP extensions, especially after go-live.
S/4HANA Extensibility and Clean Core
Deciding what must remain anchored in the system of record and what can safely move into side-by-side capabilities.
Lifecycle and Reconstructibility
Testing whether agents, workflows, extensions, context, and policies can evolve without losing reconstructibility.
Enterprise AI Governance
Putting agent governance on top of a responsibility model for truth, execution, context, lifecycle, and accountability.
The responsibility model
Core systems own business truth.
Extensions define governed capabilities.
Workflows govern repeatable execution.
Agents select and coordinate action under policy.
Assistants provide the interaction surface.
Why this is different
This is not implementation staffing. This is not generic AI advisory.
This is architectural judgment about where responsibility belongs before enterprise AI becomes ungovernable execution.
Existing BTP failure assets still matter
Post-go-live SAP BTP extension failure remains the proof layer. The same boundary discipline now applies when AI agents, workflows, extensions, automations, and S/4HANA are combined into one business action.
Boundary Model
Canonical definition of SAP BTP runtime-visible boundaries and why traditional reviews miss post-go-live failures.
Read the model →Enterprise AI Responsibility Architecture
Why agent governance is not enough without truth, execution, lifecycle, and accountability boundaries.
Read the architecture →