A good PM will go a long way to mitigating the problems with lack of generalists, but they can only go so far. The heart of the problem is that at some point,

A good PM will go a long way to mitigating the problems with lack of generalists, but they can only go so far. The heart of the problem is that at some point, you need to make decisions and solve problems where disciplines interact. In the end, you cannot really make an informed decision, or reliability solve problems in these situations without some party actually being able to grasp the cross-disciplinary picture. In the absence of actual generalists, you usually end up with ad-hoc education - crash courses where each discipline tries to educate the other (and the PM), so that they can try to get something hammered out. And this can actually be quite successful - this happens all the time, and works out very often. In fact, this is probably where many generalists (or dual-hatters) organically arise - you get a crash course in some other area while trying to figure out a problem, and then the next time you get tapped because you already have some knowledge, and you find it kinda interesting and the next thing you know, you're a generalist of sorts. I think it's quite true that in large projects you can't possibly have a generalist who actually knows -everything-. But it's also quite possible to have a good bench of dual-hatters (or dunno... 3-4 hatters), who can go a long way to relieving issues.