How does someone use a DT without some feature engineering?

I mean - I like people who say they'd try a random forest first for a given problem because it's almost always a good answer.

But yes - feature engineering is almost always the difference then.