AI for understanding and prioritizing findings

Before AI writes a single line of code, it can save the most time where developers spend it: understanding. Each S/4HANA finding points to a simplification item and an SAP Note; each legacy object has logic nobody on the current team wrote. SAP Joule for Developers, ABAP AI capabilities adds three ways to get explanations inside the tools Kenji's team already uses.

How Explain ATC findings works

In the S/4HANA system, a developer runs ATC on an object with a check variant for S/4HANA readiness (Run As, ABAP Test Cockpit With). The ATC Problems view lists the findings. Right-click a finding, choose Joule Explain, and Joule explains the simplification item and suggests a fix; you can ask follow-up questions. SAP documents the capability for findings from three checks: S/4HANA: Field Length Extensions, S/4HANA: Search for Database Operations and S/4HANA: Search for Usages and Simplified Objects, and only for simplification-item findings.

Prioritizing 3,900 findings

An explanation is only useful if it lands on the right finding. Kenji's team sorts the worklist with data the tools already provide, and uses AI to accelerate reading, not to set priorities by itself.

• Priority first: priority 1 and 2 findings before 3, as SAP recommends.
• Usage second: SUSG data in the Custom Code Migration app shows how often each object runs in EP1.
• Quick-fix availability third: the app can filter findings by quick-fix availability, which separates batch work from thinking work.
• Clean core level and process: level D objects and order-to-cash or finance objects go to senior developers and process owners.

Explaining legacy logic

The other use is the object nobody understands. Contoso's warranty accrual program, written in 2009, reads VBUK for delivery status and KONV for condition values. Before touching it, a developer asks Joule to explain what the program does, then checks the explanation against the code and against Dan's description of the month-end process. Explanations can be wrong in confident language. The habit that protects you is simple: every explanation that drives a decision gets verified against the source or a person who knows the process.

Cost and access

Explanations aren't free. SAP moved Joule for Developers, ABAP AI capabilities to usage-based billing in AI Units, so every explanation, chat and generated proposal consumes budget. That isn't a reason to avoid them; reading a simplification note and an old program can take a developer an hour, and an explanation that cuts that to ten minutes pays for itself. It is a reason to explain findings in the order the matrix sets, and not to ask for explanations of objects you'll fix with a quick fix anyway. The capability also needs setup: SAP's availability table lists the Custom Code Migration AI Assistant for Private Edition 2021 or higher, and ECC itself isn't covered, so explanations happen in the converted sandbox or CMD, not in ED1. For Contoso, that means the explanation work starts after the sandbox conversion, which is also when the remaining findings become concrete.