decision trees vs. neural networks

Decision trees/tree ensembles work well on tabular (structured) data, such as the type that you could store in a spreadsheet

Neural nets are better than decision trees for learning and making predictions for unstructured data (images, audio, speech, text, etc.), but are just as good as decision trees when using tabular data.

Decision trees are faster to train than neural networks (NNs take a long time to train)

Small decision trees are also more interpretable than neural networks; what this means is that it is relatively straightforward to understand how the tree is learning and making decisions in each iteration. Neural networks are harder to dissect in this manner; especially for large neural networks, even experts have no idea exactly how the network is coming up with an output. Note that large decision tree ensembles are similarly hard to interpret by a human.

An advantage to neural networks is that they work well with transfer learning (which works well when you have a smaller dataset).

Another advantage to neural networks is that NNs are easily scalable in the sense that when building a large ML system that requires multiple components to be strung together, it may be easier to combine NNs than it is to combine decision trees.