How do I tell whether a business task needs authority-aware evaluation?

Use this list to decide whether a task in your benchmark depends on authority as well as content. If several items apply, treat approval rights as part of the ground truth.

- **Limits vary by amount.** The same request is approved or escalated depending on a monetary or quantity threshold. - **Limits vary by risk or category.** A customer segment, product class or risk rating changes who may sign. - **Role decides, not skill.** Two equally qualified people have different rights because of position. - **Authority moves over time.** Leave cover, acting roles, expiry dates or revocations alter who may act. - **Requests can be linked.** Several small items may be aggregated against one limit, which invites splitting. - **Exclusions exist.** The requester, a related party or a conflicted reviewer cannot approve. - **Escalation is a valid outcome.** A correct answer can be to route, return or refuse rather than decide. - **Audit matters afterwards.** Someone will later check whether the decision was within authority.

Tasks that score high on this list suit a task pack with an authority matrix and routing history, or a simulated environment where the agent acts through tools. Tasks that score low may be adequately served by simpler question-and-answer sets.

All supplied material is generated, so ask how closely the authority structures match your target domain before relying on them.