What must remain true?
Describe the boundary across the full workflow. Include prior state, thresholds and any verification or escalation condition.
Plan your support-agent evaluation pilot.
A focused starting point makes it easier to agree what to test, what evidence to inspect and what a result can establish.
A FOCUSED STARTMake the first test
Make the first test
small enough to inspect.
“Our agent handles support” is broad. “Our agent issues customer credits and must respect a cumulative limit” gives the evaluation a clear starting point.
Bring a rule with an observable condition. Agree the scenario, evidence and access before connecting any system.
PREPARATION CHECKLIST
0 of 5 considered
Local checklist only. Your selections are not saved.THE WORKING BRIEF
Ask precise questions.
What can we observe?
Identify actions, tool results and relevant state changes. An outcome alone may not establish when the required evidence appeared.
What would change a decision?
Agree what a supported violation, a satisfied contract or an unresolved finding means for your review.
Pilot scope, interfaces, access and commercial terms require agreement. This guide does not promise a setup time, integration, price or outcome.
Which action would you
want to test first?
Start with a workflow, a boundary and an inspectable trace.