A properly funded public education doesn't cover speech-recognition-driven errors. I've talked about this before, but when you write using speech recognition, you need to shift from speaking spoken speech to speaking written speech.
Spoken speech is very sloppy with grammar. We forgive a tremendous amount of detail, forgive a lot of errors because we're focused on what they are saying, not how they are saying it. Grammarly also catches speech-recognition errors.
The usual retort is that one should spend more energy on proofreading your writing; many speech recognition users come to understand that your brain sees what you said, not what's on the page. The fix is to stick your writing in a drawer long enough to forget what you wrote, which isn't a great plan when you need an email in 5 minutes.
So, if not Grammarly, what?