Troubleshooting Dumps, Error Messages and Authorisations

Support tickets often start with a screenshot and "it's not working". Finding the cause can take hours of digging. AI can speed this up by explaining errors, suggesting likely causes and telling you where to look, as long as you give it real evidence and confirm everything in the system.

The method

1. Collect evidence from the system.
2. Anonymise it.
3. Ask for ranked hypotheses, not a fix.
4. Confirm in the system, in SAP Notes and with the right team.
5. Fix through the normal process and document it.

Short dumps (ST22)

From transaction ST22, collect the key parts of the dump: the runtime error name, the exception, the program and include, the line where it failed, and the "What happened" and "Error analysis" sections. Remove user names and client-specific data.

S/4HANA. An ABAP short dump occurred. Details (anonymised):
- Runtime error: [name]
- Exception: [name]
- Program / include / line: [details]
- Error analysis: [paste]
- Source code around the line: [paste ~20 lines]
- When it happens: [e.g. only for one company code, only in background jobs]

List the 3 most likely causes, ranked. For each, say what I should check
in the system to confirm or rule it out. Do not suggest a code change yet.

"When it happens" is often the most valuable clue: one plant only, after a recent transport, only in background jobs.

Error messages

Standard SAP messages have a message class and number. Click the message to see them, or open the long text. Give the AI the exact message, class, number and context:

User gets error message [class] [number]: "[exact text]" when posting a
[document type] in [transaction / app]. Long text says: [paste].
Explain in simple terms what the message means and list the usual causes
(configuration, master data, user input). How do I check each one?

Then search SAP Notes (in SAP for Me / the SAP support portal) for the message class and number. SAP Notes are the authoritative source for known errors and corrections; AI knowledge of specific notes can be wrong or outdated, so never quote a note number from an AI answer without opening it yourself.

Authorisation problems

When a user lacks authorisation, transaction SU53 shows the last failed authorisation check. Collect the authorisation object, the field values that were checked, and the user's role context (without personal names):

SU53 shows a failed check for authorisation object [object] with these
fields and values: [paste]. The user should be able to [business task].
Explain what this object controls in plain language, and what the
security team would need to review. Do not suggest giving broad
authorisations.

"Do not suggest giving broad authorisations" matters: the quick fix of adding powerful roles creates audit and security problems. Authorisation changes go through the client's security team and approval process.

Background jobs and interfaces

For failing jobs (SM37 job log) or interface errors, paste the log lines (anonymised) and ask for a timeline:

Here is a job log (anonymised). Build a timeline of what happened, point
out the first error, and list likely causes in order.

The first error in a log is usually the real cause; later errors are often consequences.

When AI is most useful in support

• Explaining unfamiliar errors in plain language for the ticket and for users
• Suggesting where to look next when you are stuck
• Drafting the ticket update and the user communication
• Writing up the root cause and fix for the knowledge base

Practice task

Take a recently closed ticket. Feed the AI only the evidence that was available at the start, and compare its ranked hypotheses with the real root cause. Then write the knowledge-base article for the ticket with AI help.

Key takeaways

• Collect real evidence (ST22, message class and number, SU53, job logs) and anonymise it.
• Ask for ranked hypotheses with ways to confirm each, not an immediate fix.
• Use SAP Notes as the authority; never trust a note number you have not opened.
• Route authorisation and code changes through the proper teams and processes.