I worked at a small robotics company in the Netherlands, technical university spinoff. We did tests in care homes for elderly and disabled people and we actually had clients that wanted to use our robots and were very positive about them.

But we were just too early. The robot didn't even need to be very autonomous, shared autonomy was part of the plan (fake it till you make it and honest about it :-) ). I controlled the robot from the other side of the (small) country to give people their teddy bears and TV remotes etc.

Problem was that we tried to do our hardware completely by ourselves, whereas nowadays there are several options for mobile manipulator robots that look good enough to put in a home. We spent more time getting the robot to work than on useful applications (we had dozens of activities of daily live a robot could support). And even the getting our robot to work could be done more efficiency (in hindsight). We should have used e.g. an EtherCAT motor controller instead of writing of doing it all ourselves on a micro-controller.

We didn't progress enough quickly enough for our investors and subsidizers and ran out of money.

A another great lesson was this though: when you hear talk about robots in care, the complaint I get a lot is that that is sad for the clients, because they don't get someone to talk to (but a robot) and they will get lonely. That turns out not entirely the case. A lot of disabled clients told me they don't want to talk to the care-givers all that much, they just want their drink/toilet break/pen they dropped etc. They have friends and family and don't feel lonely. The complaint about loneliness is coming from the care-givers. For elderly people, it may be different though, but I found that very interesting.