There are so many problems with human mods. For one, they have lives and sleep. A community server should ideally be run 24/7 (because otherwise finding new servers all the time sucks and massively degrades the experience relative to the convenience of matchmaking), but when there are no mods around, the server is liable to be ruined. This can lead to the long-term stagnation or even death of the server.
And human mods suck. Really, really suck. They don't have perfect information, and their false positive / negative rate is probably an order of magnitude worse than an anti-cheat. Even a perfectly neutral moderator is bad at judging. And perfectly neutral moderators don't exist. It was not remotely uncommon to be banned from a server for killing the admin too many times.
This is exacerbated by the fact that moderating itself sucks. Moderators are there to play the game. Actually moderating on the side of that is a burden. This leads to the same incentive structure that you see all over the internet, where volunteer moderation mostly attracts people who are interested in power tripping, because otherwise there's very little appeal, and bad actors are more common than people willing to do it out of the goodness of their heart.
There are great points about private server browsers, but it's also just a massive pain to find servers that both have a gameplay configuration you like and also have decent moderation. Improving anti-cheat systems was the obvious way forward rather than relying on humans who are much more fallible. Kernel anti-cheat should fortunately only be a stopgap solution and not something that is here to stay in the industry. Kernel anti-cheat was a result of the failure of Windows to provide sufficient security features themselves. Riot's Vanguard doesn't rely on kernel access for macOS, and Microsoft is actively working on improving the kernel security such that it shouldn't be necessary on Windows in the future either. With any luck we'll be able to forget this era of anti-cheat in a few years.