No, I haven’t done it. LLMs did not exist when I last worked a proposal. (Tell me how old you are . . .)
People can learn by doing, and one good way to illustrate how to write well is to expose writers and editors to bad proposal responses and to get them to improve the writing. Here’s the problem. You can’t use a bad example within a company without potentially humiliating the author. Neither can you use bad examples across companies, for confidentiality reasons. What you want is bad text on a relevant topic, created just for training purposes.
However, I have tried to write bad text and it’s a difficult and time-consuming task. Given all the bad writing in Proposal Land, you might think it would be easy to craft a bad response, but it is not so: Gobbledygook won’t work. The text has to make sense and be largely correct from a technical point-of-view, but also be badly organized, repetitive, and full of believable English errors. You want to approximate the mistakes that proposal newbies make.
So. If you would like to have bad text for training purposes, consider using an LLM to draft and revise it. As part of your analysis, check out this post by Scientist Sees Squirrel, and learn from his well-documented experience and results–albeit in a different field of writing.
The tl;dr: it’s possible to use LLMs spectacularly badly, both unethically and ineffectively, either for writing or for mentoring it.
It’s also possible to use LLMs well – in ways that help writers think about writing, rather than in ways that let them avoid thinking about writing.