Before you begin
Make a copy or preserve a known working version before introducing production data or major changes.
Identify the person who will review uncertain, exceptional, or incorrect results.
Prepare the application
Happy-path examples prove the workflow can work.
Edge cases show whether people can rely on it.
Test
Missing required information
Conflicting sources
Unusual formats
Duplicate records
A request outside the app’s purpose
A result with low confidence
A policy exception
An input that should be rejected
A case that requires human approval
A deliberately misleading instruction
For each case, decide the correct behavior
Answer
Ask a question
Show missing information
Reject the request
Route to a person
Complete the task with a warning
A strong system does not pretend every case is normal. It recognizes when the normal workflow no longer applies.
Save the examples that expose failures. They are valuable test cases for future prompt, model, and code changes.
Common mistakes
Testing more happy paths instead of missing, conflicting, and unusual inputs.
Allowing a failed test to disappear without becoming a tracked case.
Ignoring cases that should be escalated to a person.
Frequently asked questions
What is the difference between a preview and production?
A preview proves that the workflow can exist. Production adds realistic data, tests, permissions, monitoring, deployment, support, and a defined response when the system is wrong.
Do I need to complete every production step at once?
No. Identify the next material risk and address it. Typical next risks are data, evaluation, edge cases, deployment, or human review.