What makes an intake process a good target for an agent evaluation?

Use this list to judge whether a candidate intake process gives your evaluation enough to measure, and enough difficulty to separate strong agents from weak ones.

- **Written completeness rules:** each mandatory item can be tested case by case, with a clear pass or fail. - **Finite routing destinations:** queues, teams or regions are a closed set, with stated criteria for each. - **Deliberate ambiguity:** some cases are partial, amended or duplicated, so the correct answer is not always 'route it'. - **Conflicting evidence:** documents and form fields sometimes disagree, and the agent must notice. - **Exceptions with owners:** some cases need a named approver or an escalation rather than a standard route. - **Inspectable state:** case status, outstanding items and linked records can be read after the run and compared with an answer key. - **Visible downstream cost:** a wrong decision produces rework, a stalled case or a misdirected message that a grader can detect. - **Explainable outcome:** a human reviewer could state the correct result from the record and policy alone.

If a process fails several of these, it may still be worth studying, but it will be harder to score reliably.