no broadly capitalist system can really produce a society that works well for anyone other than the favoured few.

It's worse than this. Capitalism only even works for the particular capitalists on a temporary basis. As the system works, the class of people owning any capital at all is narrowed, which Marx called "proletarianization" (to wit: owners of small shops are pushed out by, say, Wal-Mart, and thus forced to become wage-laborers instead). To the extent that capitalism is doing something decent for humans, this is fine; to the extent capitalism is just a big societal paper-clipper, we vitally need to stop it.

just as with the prospect of AI, I suggest that the question to ask might not be "how can we defeat this powerful hostile thing?" but "how can we stop this powerful thing being hostile and make it act for our benefit?".

I disagree slightly: AI is far more beneficial than capitalism ;-). AI is a blank optimization process, so to speak, an AIXI implementation doesn't come with an unfriendly utility function built-in, you have to add one and then unleash it yourself. With AI there at least exists the possibility of specifying Friendliness and getting it right; the "evil by default" behavior isn't some cosmic force that hates us, it's just that we only like a tiny subset of possible universes.

Not least because the actually-existing societies that seem to do best at providing a decent life for most of their people are broadly capitalist, with various mechanisms attached that try to fix raw capitalism's tendency to screw over all but the wealthiest.

I think we need to speak seriously about the varieties of social optimization processes.

I agree that social democracy has produced the best observed results for human beings, and in fact contains explicit mechanisms (ie: democratic participation in allocation of the means of production) for ensuring human-friendliness. I also wish to note that markets should be considered a mere component of capitalism, and that pro-capitalist ideologues who focus primarily on markets are, essentially, rather like mechanics who focus on getting the greatest horsepower rather than on what the engine is for. So we can assume that the category of human-friendly economic systems contains at least social democracy, and also probably cooperative-based market systems, syndicalism, possibly democratic state-socialism, commons trust-based systems, and many others.

However, there are many variables that go into a market system. What we currently have is neoliberal financialized capitalism: liberalism level = state compels labor but subsidizes capital accumulation, definition of money to be accumulated = virtual credit-money rather than industrial production, allocation of the means of production = private profit-maximizing companies, donation-funded NGOs, state agencies, and almost nothing else.

Slavoj Zizek said in 2011, "It is easier to imagine the end of the world than the end of capitalism". David Graeber has characterized neoliberalism as an ideology which, when given the choice between making capitalism seem the only possible economic system and making capitalism a long-term viable economic system, will always choose the former. It is not enough to simply rail on behalf of "free markets" or "workers' revolutions", we need to be rationally unpacking the variables that make social optimization processes function one way or another, and determining what values for those parameters result in something we actually like.

It is not enough to scream, "Another world is possible!" Dozens of other worlds are possible, which is why it's imperative to stop pretending that this is the only or ideal world despite the fact that so many don't like living here.