Wouldn't the sender be the better judge of the accuracy of what they wanted to communicate? If they feel the LLM version more closely matches their intent, then we should accept it.
There are good arguments to get to know someone "mistakes and all", I just don't think this is a particularly good one. No matter how much you (think) you know someone, they probably know them(selves) better.
The intent behind these mentions was to "promote the use of Copilot" and "showcase Copilot usage scenarios in a variety of different ways".
These are quotes from resolved comments in the review to a similar but slightly earlier PR[1], from the same author, to the one that introduced the specific mention referred to in the issue[2].
And that's why these suggestions to use Copilot probably don't belong in the docs: their intent was to promote a product.
(To be fair, in the first comment quoted above the reviewer asks that the Copilot section be moved to the end of the document, prioritizing teaching the user about the actual feature the article was about).
One additional problem, already pointed out by others, is that Copilot seems to be the only AI tool showcased in the docs. Besides suggesting the intent of these suggestions is purely promotional, this reflects poorly on an organization that should be independent from Microsoft.
A complementary method is to attach a writable NFC sticker tag to the phone. Though it has to be placed far enough from the phone's NFC antenna in order for both/either to work.
The upside is that you get a second tag of your choice (physically) on your phone. There are even UID-writable sticker tags out there (even if they can be a tiny bit harder to find).
You also don't need to replace your default transit card, which could be inconvenient depending on where you live.
There are good arguments to get to know someone "mistakes and all", I just don't think this is a particularly good one. No matter how much you (think) you know someone, they probably know them(selves) better.