Configuring anomaly detection and alerts

8 min

Enable anomaly detection

Navigate to Intelligence → Anomaly Detection and toggle the feature on. Jarvis will begin building a baseline of your normal transaction patterns — this takes 2–4 weeks of transaction history to calibrate effectively. You can also import historical data to accelerate baseline training.

Choose detection sensitivity

Select from three sensitivity levels: Conservative (only flags clear outliers — 3+ standard deviations), Standard (our recommended default — catches most meaningful anomalies with minimal noise), and Aggressive (flags anything unusual — best for businesses with very consistent spending patterns). You can adjust per-category if needed.

Set up notification routing

Under Settings → Notifications, configure who receives anomaly alerts and how. Options include email (immediate or digest), in-app notification, and webhook to Slack or your incident management tool. For larger teams, route different anomaly types to different people — finance anomalies to your controller, expense anomalies to your department heads.

Review flagged anomalies

Anomalies appear in the Alerts feed with the detected transaction, why it was flagged (e.g., '340% above your 90-day average for this vendor'), and quick action options: Mark as expected (removes from alert queue), Investigate (opens transaction detail with context), or Create rule (prevents this pattern from being flagged in future).