Canadian and U.S. case-outcome predictions with the rigor of a senior associate's review. Transparent methodology. Free predictions for self-represented litigants in Canadian refugee, disability, and judicial review matters.
Choose your jurisdiction to see available tools and pricing.
Paste the relevant facts from your case — a lower decision, the appellant's position, key evidence, legal issues. In about 30 seconds, NorthLaw returns an outcome probability with a real confidence interval.
Every prediction shows its confidence tier and the historical accuracy of that tier. If the model isn't sure, it tells you.
Three commitments that distinguish NorthLaw from legal AI that produces confident-sounding predictions without showing its work.
When we say 68% ± 12%, those numbers come from 50 independently trained models, not marketing rounding. Every prediction shows its true uncertainty.
Training data size, validation method, AUC score, accuracy by confidence tier — all public, per-tool. No proprietary black boxes. No numbers we won't stand behind.
The three A2J tools — RAD, SST, and FC-JR — are free forever. No signup. No time limit. Predictions shouldn't be locked behind a paywall when someone's refugee claim is on the line.
No per-seat fees. No feature tiers. No surprise charges.
Cancel anytime. Annual billing saves ~17%. No hidden fees.
NorthLaw was built by Karim Souidi, a senior statistician based in Toronto.
I'm not a lawyer. I spent 11 years at Legal Aid Ontario — first as Statistician, then as Senior Consultant in Statistics & Analytics — where I designed the statistical systems (forecasting, intake automation, efficiency measurement, and survey analytics) that LAO used to allocate resources and serve clients.
Before that: Fulbright scholar at UC Berkeley, 11 years teaching Statistics and Machine Learning at Canadian universities (Ryerson, McMaster, Sheridan), Master's in Econometrics. Today I continue to work in government analytics and public-sector methodology.
I built NorthLaw because legal predictions should come with the same honesty we expect from any other statistical forecast — published accuracy, real confidence intervals, and a clear signal when the model is uncertain.
Every model on NorthLaw publishes its accuracy per-tool, shows calibrated confidence intervals from real bootstrap resampling, and is trained on real court decisions. If we don't know, we say so.
Try the free tools — no signup, no credit card, no catch.
Start with a free prediction →