But there's definitely uses for a tool that produces a lot of boilerplate content that a human can then proofread (ex: writing cover letters, filling out unstructured paperwork, etc.)
Edit: I personally underestimated the use-case, because the "more optimal" thing to do when auto-filling out paperwork would be to define a formal data-structure with annotations for each field, and then have a system read annotations and spit out autosuggestions. But this is a lot of overhead.
If you can have a system that takes in arbitrary text, parses it, and fills it out with pretty good guesses, then that can be much more viable than the previous approach, as it doesn't require an overhaul of, say, the way the US bureaucracy generates paperwork. It's all on the consumer-end, the producer doesn't need to cooperate.