I have personal experience in customer service scenarios where people classed by 'the computer' with an LTV score below a certain value were not to be retained if they wanted to cancel our service (c.f. for a higher level LTV customer we might offer a promotion or something) and people with a score even lower were to be actively encouraged to leave if they reported any problems - even to the point where we were supposed to suggest they try somebody else rather than spending any time fixing their problems.
Granted, these were (allegedly) people who were actively costing the company money in that it cost more to service their account than the revenue they provided, but it was an automated decision made by an algorithm that the phone and retail reps just had to enforce.
Depending on that score they present more or less dark UI patterns to that user. Past a certain threshold they probably don't want to keep the insuree anyway, so if they switch it's still good for the insurer. They can calculate the likelihood the user will switch and how much the insurer will lose in premium, versus how much they'll win if they delay claims as much as possible.