100% local · free
Document → data
A scanned invoice becomes structured, cross-checked data in seconds, without leaving your browser.
Launch the demoORIA Lab · Weak signals
300 fictional customers, 18 months of history. The detector combines trend, change detection and peer comparison: tune its sensitivity and compare it with a classic threshold rule.
How it works
The slope of activity over recent months, computed by least squares and expressed in % per month.
A cumulative-sum test (CUSUM) spots a lasting slide below the customer’s usual level, even a slow one.
A robust score (median, median absolute deviation) positions the customer’s trajectory against its segment.
These indicators, with payment delays and complaints, form a risk score; the alert threshold is the slider you set.
In production, signals are built with your teams from your own data (transactions, usage, delays, complaints), thresholds are calibrated on your past cases, and every alert is explained and checked against real cases before rollout. Models can be richer; the decision stays human.
Weak-signal detection & decision supportOther demos
100% local · free
A scanned invoice becomes structured, cross-checked data in seconds, without leaving your browser.
Launch the demo100% local · free
Ask a question about internal procedures: the answer always cites its source document, or is not given at all.
Launch the demoNext step
A structured conversation with an AI architect — no commitment, no jargon.