The goal with most of these AI features is not to solve a real problem users are having, it's to add a feature that uses AI. This will not change because it's not wrong of the individuals making the decision. The project manager gets to say he shipped a cutting-edge AI project. The developers all get to put experience working with very hireable technologies at a serious company on their resume. There will be no adverse impact to the bottom line, because the cost to develop the shitty AI feature is a drop in the bucket, and the cost to create a competing product that accomplishes the core thing users are using that product for but without feature bloat would be very high, and probably unsuccessful since "less feature bloat" has never been sufficient to break the static friction threshold for users to switch.

So it won't change, because there is no lesson to learn. No individual involved acted irrationally.