I would say, find the people who are most willing to listen and try with them first. Especially ones who are in the middle of having a grievance with the system, find ways to tie it into an explanation about why those things occur. Under some circumstances this could be manipulative (playing off of people's upset), but from the ML standpoint, the goal is to give them real theory and practice of solutions that can help them. Don't get lost in the weeds of thinking they all have to agree with you theoretically if they aren't likely to put it into practice anyway. Like... wages are shit, can a labor union be formed? Dealing with obnoxious landlords, can a tenants union be formed? I know unions alone are not ML (and it's just an example), but things like that may lead to them being more open to listening on ML theory.

Cold opening on ML talking points can be a lot, especially for somebody who is really immersed in thinking "communism bogeyman scary." So try to meet them where they're at first and go from there.

And in general, try not to take it personally if you struggle to get through sometimes. In my experience, sometimes there are situations where one person is simply better at connecting with another and getting them to listen. Perhaps because their minds work similarly, I'm not entirely sure. Like I am good at talking in the abstract, on concepts, but remembering specific dates and events to cite is hard. This is not so much a problem if I'm talking to someone else who is similar because they may not find the events citing stuff all that moving anyway. But if someone does more think in those tidbit "trivia" events, it may be a person who can rattle off events is going to stick with them a lot more easily. This is a very general example and it is probably going to be a lot more nuanced in most situations, but something to think about.