DeepL might not (yet) be as good as a good professional human translator, but I'd say it's already better than the average professional human translator. We have a contractor who translates technical texts from German to English and DeepL makes less mistakes and even gets some industry-specific terms right, whereas the human translator with no domain knowledge may correctly translate the term, but chooses a different one that immediately sticks out like sore thumb to anyone who works in the field. I'd rather give a text to DeepL and fix a few things than give it to contractor and still have to fix a few things.