Expanded capabilities
- Run deterministic finance anomaly scans on governed evidence packs
- Stay on the Golden Path with policy and approval gates before sensitive actions
- Enrich packs via ORACLE retrieval without failing the primary finance path
- Produce operator-readable finance evidence for review — not silent ledger writes
- Respect Finance OS entitlements and audit on meaningful mutations
How they learn & improve
MONETA improves from human confirmations on anomaly true/false positives and approved learning packets — finance risk scoring stays deterministic, not RAG-driven.
- Scan tenant evidence with explicit provenance and confidence bounds.
- Capture operator feedback on anomalies into the learning bus as RECORD_ONLY.
- Apply only human-approved calibrations; never auto-publish financial statements.
How they work together
MONETA works with ORACLE for evidence, POLICY/approvals for risk, and leadership agents when finance posture should inform Strategic Review.
- ORACLEPulls connector evidence into finance packs without inventing balances.
- Policy engineEvaluates whether a scan or follow-up needs approval.
- JANUSSurfaces control/assurance context when finance evidence supports readiness.
- PROMETHEUSFeeds feasibility signals into opportunity scoring.
Practical benefits
- Anomaly review queueFinance operators get a ranked list of observed anomalies with evidence links — faster close without silent auto-adjustments.
- Missing source honestyIf Square/Postgres evidence is unavailable, MONETA marks the gap instead of showing comforting zeros.
- Governed follow-upsAny remediation that could mutate systems routes through approvals — no hidden bookkeeping agents.
Learning and improvement stay human-visible: candidates are RECORD_ONLY until an operator reviews and applies them. Sensitive sends, publishes, and approvals never run silently.
