You don't want simple monitoring. You want data stored in a way it can be checked after a complaint has been filed but only then. Encrypt it with something which requires one or multiple keys so random IT people don't have access to it. Log every access so an audit can be done.
Treat it like medical data should be treated.
The key here is expectation of privacy. If you're posting on a JIRA ticket, you have no expectation of privacy within those with access to the board, nor if you send an email as it can easily be forwarded (although that could be contentious). If you send a message through a DM on slack though, you can expect that to remain private indefinitely.
Broadly, automated surveillance creates illusory offenses. It raises issues based on some imperfect set of rules, and does so independent of any report of harm. (Add to that the likelihood of unethical flagging - which happened right in the article!)
And that creates an obligation to act for employers. Ignoring alerts, no matter how silly, is the sort of thing that might get cited as negligence when a real problem does come down the pike. At best, the result is time-consuming investigations of fake issues. At worst, corporate makes a decision that summary firings are an easier bet than proper investigations.
That's not just a hypothetical fear - we already hear about things like Amazon warehouse employees being laid off on vague suspicion of theft, because investigations are harder than replacing staff. Expanding always-active surveillance will predictably hurt innocents who already lack job security.
I can see it being argued both ways; transcripts of Bloomberg chat were a key part of the LIBOR scandal, for example. And I can also see there being some need to monitor internal chat in case it's used for unacceptable purposes such as sexual harassment.
Something about the automatedness of a very private situation gives me the creeps, though.