Situations let you describe cases, run Scout Agent and inspect its behavior. The product identifies these runs as sandboxed so nothing real gets sent. Use them to test both success and exception paths.
Before you start
You need access to Situations and the referenced records. A situation that mentions a deleted record may be unavailable.
Compare unconfirmed, partly confirmed and unanswered purchase order situations before running a test.
Build and review a test set
Open Scout → Test → Situations.
Choose Add situation or start from samples. Give the case a task title and description; mention relevant records where supported.
Describe what a good result would look like, including important constraints or a condition that should require approval.
Select the situation and run it. Open the run to review the plan, tool calls, questions, proposals and completion state.
Rate the result Good, Acceptable or Poor and leave a specific note.
Adjust the affected knowledge, guidance, playbook or permission, then rerun the relevant cases.
Check the result
The run shows how Scout Agent handled the situation, and the rating captures your assessment of quality. A run that stops at a question or proposal can still demonstrate the intended approval boundary.
Use a balanced set of cases
Include a straightforward order confirmation, a missing contact, an ambiguous supplier reply, a permission boundary and an outcome that cannot be achieved. These test different parts of the setup.
Compare the run’s quality with its turns and tool calls. A short run that reaches the wrong result is not an improvement.
What a test does not prove
Sandbox behavior helps evaluate decisions. It does not establish that a real supplier received email, a provider accepted a calendar change or a device received a notification. Verify those effects in the appropriate real workflow before relying on them.
If something does not look right
What you see | What to check |
The situation will not run | Check referenced records, current run state and the visible error. |
A rating is unavailable | Run the situation first and inspect the completed attempt before grading it. |
Should I rerun every situation after any change?
Should I rerun every situation after any change?
Rerun the cases affected by the change, including a nearby case that should keep its previous behavior.
Continue your work
Open the relevant area in Cartograph to continue with your workspace data.