Resources — Field notes

The eleven-month vendor

2 min read

The vulnerabilities that hurt you most are rarely the obvious ones. It is the vendor behavior hidden in historical data that looked ordinary for eleven months before AI transaction monitoring flagged it as an event requiring action.

Static rules cannot see this. Eleven months of ordinary is exactly what fixed thresholds are tuned to ignore. Behavioral baselines, by contrast, learn what normal looks like per vendor, account, and process: typical volumes, values, frequencies, and timing. Activity that deviates well beyond baseline is flagged even when no static rule exists for it.

History as the lens

Live and historical data combined is the whole point. AFM ingests scheduled feeds of audit and transaction records, maps them to a common model, and runs detection in layers: rules for the known, baselines for the drift, AI for the relationships, scoring for the priority.

Baselines update as business patterns evolve, which keeps alerting calibrated instead of drifting into noise. Ordinary stays ordinary; the eleventh month does not.

All blog posts

See these patterns in your own data.

See the platform work against a real slice of your own data.