Agent #010 · Reporting & Revenue Insights
Weekly Revenue Brief Agent: governance & failure modes
What this agent is built not to do, and the specific ways it can fail if misconfigured.
Governance notes
- Never invents a metric — the defining rule of this agent. Every number traces to a source system or an upstream agent’s output.
- A section with no available source renders as explicitly unavailable, never as a guessed or extrapolated figure.
- Writes only to the published brief document — it doesn’t write back to any CRM record.
- The "questions for leadership" section is a structural requirement, not a nice-to-have, forcing risks to surface instead of going unspoken.
Failure modes
- A metric gets fabricated to fill a blank section: prevented by requiring a source citation per figure in the generation prompt, verified by a QA test using intentionally missing source data that must render as "unavailable," not a number.
- The brief goes stale or fails to publish: a delivery-confirmation check in the workflow alerts the owner if publishing fails.
- The questions-for-leadership section is empty filler every week: caught by a manual QA read for genuine substance, not just a check that the section exists.
- A figure gets attributed to the wrong source in a multi-agent merge: every populated figure includes its source agent or system inline, so misattribution is visible on sight.