The problem is that every facet they add to the analysis increases the complexity, making it harder to price risk - meaning they'll eagerly overprice [0] (based on "what if") with some idea that "the market" (ie nobody) will sort it out. This creates a race to the bottom against anything that might seem "weird", aka regular people just living their lives (while being put under a microscope), similar to how the ever growing housing bubble has painted most houses beige.

Your bushes only seem like a reasonable example because you are looking at them in isolation - it was only a single issue you had to deal with, and something you seemingly wanted to deal with anyway. If they had instead blasted you with a litany of different issues, or a bush that you wanted to keep for sentimental reasons, or you simply didn't have the time/resources to create your own counter-documentation and operate their heavyweight bureaucracy, you'd be singing a much different tune.

And while these things can happen anyway with regular in-person home inspections, the point is that increasing surveillance and unaccountable "AI" make it much easier to bury customers in a deluge of complexity making for even-less-competitive markets.

(I would give a healthy list of examples of my own home being deficient in many ways an insurance inspector would call out yet are in the process of being managed, but I'm sure it would just invite a lot of "well ackshually that's dangerous and weird!" out of touch responses from people whose first instinct is to call someone rather than to fix something themselves)

[0] Notice how they weren't going to raise your premiums by $30 a year or whatever, but outright cancel your policy - in other words a massive overweighting of the actual risk from some bushes growing into the house.