Custom SAP code often represents years of business decisions. Some extensions remain essential, some can be replaced and some may no longer be used. The difficult work is establishing which is which without disrupting the processes that depend on them.
AI can assist with reading and explaining code, preparing documentation and proposing changes. Those proposals still need evidence from the actual landscape, supported analysis tools and qualified review.
Begin with an inventory you can inspect
Identify custom objects, their owners and their role in business processes. Include interfaces, scheduled jobs, reports, enhancements and external dependencies. Repository code is one source of information, not the whole operational picture.
Usage data helps distinguish active objects from candidates for retirement. Its coverage matters: a short observation period may miss a year-end job or an infrequent exception process. “Not observed” is not the same as “unused”.
Record the uncertainty instead of asking a model to resolve it from naming patterns alone. An object with a confusing name can still support a critical business task.
Use AI for bounded analysis
An assistant can summarise a routine, identify potential dependencies in the material supplied to it or draft an explanation for a developer. Give it the relevant source and a clear question.
Check the explanation against the code and the system. A model can misread dynamic behaviour, configuration-driven paths or an interface contract. Its generated description is a working aid rather than authoritative documentation until reviewed.
Preserve references to the objects and versions used. If the source changes, an old explanation may no longer describe the current implementation.
Combine static findings with business evidence
SAP’s Custom Code Migration documentation describes analysis findings for supported readiness checks. Use the tooling and check variants appropriate to the actual system and target release.
ATC findings can identify technical issues that need attention. They do not establish that a proposed replacement preserves every business rule. Ask the process owner and experienced developers to review behaviours that are not captured in the available tests.
Keep the migration target explicit. A recommendation suitable for one SAP release or deployment model may not apply unchanged to another.
Review changes as software changes
A suggested refactor should be small enough to inspect. Compare the original and proposed behaviour, run the relevant checks and test representative inputs and exceptions.
Include integration behaviour, authorisation checks and performance where the change can affect them. Syntax correctness is not sufficient evidence that financial or operational logic remains correct.
Transport and deployment should follow the organisation’s existing change process. Do not give a code-generation assistant unrestricted production access because it produced a convincing explanation.
Protect source and business information
Source code can contain proprietary logic, identifiers and sensitive configuration. Review how a model provider, retrieval index and diagnostic tooling process it. Keep secrets out of prompts and apply access controls to derived documentation.
Private hosting can be an option, but it does not automatically resolve every data or security requirement. The complete data flow and operational arrangement need review.
Prioritise with explicit decisions
For each candidate object, record whether it should remain, be adapted, be replaced or be investigated further. Attach the supporting usage, dependency and test evidence, along with the responsible owner.
Avoid a blanket claim that custom code is technical debt. The useful question is whether a particular implementation creates maintenance cost, upgrade difficulty or operational risk relative to the alternatives available.
Ryzolv’s AI Software Engineering work can support bounded analysis and development tasks. AI assistance should make the evidence easier for people to inspect, while responsibility for the business change remains clear.
Written by Anshul Kaushik, Managing Director at Ryzolv.
Back to Insights