FP&A Human Review Ledger
Review gates, overrides, approval chain, and replayable human governance for forecast decisions
Live trust drill
Human review ledger
Show governance and override
Human Review Ledger
This page makes the human-in-the-loop framework concrete. It shows who reviewed the forecast, what they changed, why they changed it, and how the decision can be replayed later.
Forecast approval chain
1
Forecast generated
Agent proposes narrative, range, and confidence.
2
Analyst reviewed
Analyst validated drivers and requested one rerun.
3
Director approved
Confidence floor satisfied after external-signal check.
4
CFO accepted
Forecast approved for executive conversation.
Recorded actions
| Role | Action | Reason |
|---|---|---|
| Analyst | Modify assumption | FX headwind updated from latest treasury feed. |
| Team lead | Request rerun | Commodity shock range needed wider confidence band. |
| Finance director | Approve | Confidence restored above governed floor. |
| CFO | Accept | Forecast fit for board and capital dialogue. |
Override discipline
- Every accept, reject, modify, rerun, and override action is logged.
- Evidence pack, scenario pack, and policy pack remain linked to the final decision.
- The audit replay must show both agent reasoning and human judgment, not only the final answer.
Trust rule: bp Sphere should never hide disagreement. It should preserve challenge, override reason, and final accountable approver.