Why it matters
ML has quietly run business for years: spam filters, fraud checks, recommendations, ad targeting. Understanding the principle (patterns from data, with data quality setting the ceiling) inoculates you against both the hype and the fear when every vendor rebrands as AI.
How it works
Models are trained on historical examples until they generalise to new ones: these transactions were fraud, so flag ones like them. Predictions are probabilistic (confidence, not certainty) and models drift as reality changes, which is why they need monitoring, not just launching.
What to do about it
When a vendor says "AI-powered", ask what's learned from what data, and how it's kept current. The answers separate products with real machinery from products with a marketing layer.
