Which scenario is a good use case for machine learning?

Classifying data with no prior examples describes zero-shot or unsupervised scenarios where ML has limited effectiveness without labeled training data.

Tasks requiring human experience and intuition, such as nuanced judgment calls, are areas where ML currently struggles compared to humans.

Recommendation systems use historical interaction data - such as purchases, clicks, or ratings - to train models that predict items a customer is likely to prefer. This is a well-established ML pattern with abundant labeled training examples and a clear objective function, making it highly suitable for machine learning.

Ethical dilemmas involve value judgments and moral reasoning that are outside the scope of ML model capabilities.