AI-Assisted SAP Modernisation with a Clear Review Path

SAP modernisation involves more than translating old code. Teams need to decide which business behaviours remain necessary, which extensions should change and how the new landscape will be tested and operated.

AI can help prepare analysis and proposed code changes. It does not decide whether an extension belongs in the core, establish upgrade compatibility or approve a business-process change.

Define the target landscape

Document the current products, versions, deployment model and planned destination. Identify the business processes and interfaces included in the change.

Maintenance and upgrade decisions should use SAP’s current product-specific information and the organisation’s contracts. Avoid applying a single general deadline to every system. SAP’s maintenance strategy page is a primary starting point for the relevant product scope.

The target determines which tools, interfaces and extension options are available. A recommendation without that context can be technically plausible and still unsuitable.

Treat clean core as an architectural decision

A clean-core approach aims to keep extensions and changes within supported, maintainable boundaries. It does not mean that every organisation must remove every custom behaviour.

Review released interfaces and the supported extensibility options for the target. SAP’s ABAP Cloud extensibility documentation explains the relevant development model. Availability and suitability still depend on the actual landscape.

For each extension, ask what business need it serves, whether standard functionality now addresses that need and where the remaining logic should live. Record the decision and the evidence behind it.

Give the assistant a bounded job

Useful tasks include explaining a selected routine, drafting tests, preparing documentation or proposing an adaptation for a known finding. Supply the relevant source, target constraints and expected behaviour.

Require references to the code or documentation used. Review generated statements about interfaces and compatibility against the target’s official documentation rather than accepting a model’s recollection.

An assistant should not declare an object unused from code alone. Combine source analysis with representative usage evidence and process-owner input, including infrequent but important work.

Build a testable change set

Keep proposed changes small enough for qualified review. Run the supported static checks and relevant tests, then verify the business behaviour and connected interfaces affected by the change.

Test negative cases as well as normal transactions. Include authorisation failures, missing data and the exceptions the current process handles. Preserve existing controls while checking whether the new design introduces different failure paths.

SAP’s conversion and upgrade guidance describes ATC and other checks in supported scenarios. Passing a check is evidence for its scope, not a general certificate of business correctness.

Keep access and release authority separate

Code and business data supplied to an AI tool need approved handling arrangements. Review model processing, retrieval stores and logs. Restrict tool credentials and do not expose production secrets through generated prompts or documentation.

A developer or authorised reviewer approves the change through the organisation’s normal process. Model-generated code should not bypass transport, testing or segregation-of-duties requirements.

If a deployment fails, define how the team restores service and reconciles affected records. A rollback of source code may not reverse every data change.

Measure the actual contribution

Track the work AI helped prepare and the effort needed to review and correct it. Do not assume a fixed percentage improvement in delivery time. The result depends on code quality, test coverage, landscape complexity and the task assigned to the assistant.

The objective is a maintainable change supported by evidence. Our AI Software Engineering capability fits that surrounding development and integration work, with professional responsibility for the SAP landscape remaining with the appropriate team.