Supervised learning is training with labeled examples: each piece of input data comes with the correct answer attached, like a photo labeled "dog" or a house listing labeled with its actual sale price. The model's whole job during training is to get better at predicting that known answer, which makes supervised learning the go-to approach for classification and prediction tasks.
Unsupervised learning works with data that has no labels at all. Instead of predicting a known answer, the model looks for structure on its own — grouping similar customers together, or noticing that certain transactions look unusual compared to the rest. It's the right tool when you don't already know what pattern you're looking for.
Many real systems mix both: unsupervised techniques to explore and organize raw data, then supervised techniques once a labeled dataset can be built from what was discovered.