My last startup was (yet another) attempt to solve unemployment. As we got closer to the problem with our market research, built an initial product, and tried to sell it, we realized that the field is...well, let's call it "fractally fucked up". The job/employment market is a mess of millions of different actors, all of whom are doing the wrong thing because their incentives are perverse, they lack full information, and they wouldn't act on that information because they don't trust it even if they had it. You can't solve this problem by throwing around money, because you have to change behavior and you only change behavior through long-term, deliberate attention on the part of all of the individual actors.

I'd posit that all of the "ugly problems" of this world - unemployment, education, inequality, health care, human rights - result from similar fractally fucked-up systems. It is far, far easier to solve the problem for yourself and a small number of other people than to solve it for everyone. In some cases - like inequality or power dynamics - it's not possible to solve it for everyone, because the definition of the problem changes as you make progress.