HomeAIAnomalies
AI Anomalies

Catch what you miss.

LLM-powered anomaly detection on your transaction stream. Surfaces duplicate charges, weird vendors, and amount outliers — the stuff you'd miss in a 200-row statement.

Get protected See providers
What you get
Duplicate detection
Same vendor + same amount + within 24h? Flagged. We've caught $89 streaming subscriptions double-charged.
Amount outliers
A $1,200 charge when your median is $40? Flagged with the AI's confidence score (95/100 = high).
Vendor anomalies
New vendor you've never paid before? Different MCC code from the same vendor? Surfaced as a "look here" item.
LLM-verified
Pattern detection is rule-based (fast). LLM verifies the top 5 daily and explains why it's suspicious. Human-readable reason.
Email + in-app alerts
High-confidence anomalies ping you via email + in-app notification. Mid-confidence stays in your Anomalies inbox.
Dispute in 1 click
Click "Report" on an anomaly → dispute workflow opens. We acknowledge in 24h, file with scheme in 5d.
In practice
AI inside · Marcus · got pinged on a duplicate
Marcus's $9.99 Spotify was double-charged due to a billing glitch. The Anomalies feed caught it in 3 hours, AI confidence 97/100. He reported, got refunded in 5 days.
3 hours
Detection to alert
At a glance
97/100
High-confidence threshold
24h
Dispute ack SLA
5 days
Scheme filing
10 days
Provisional credit