In my experience, Haskell takes a bit more to "get." If Java is good enough and managers are able to find multiple, interchangeable developers, Haskell doesn't make a lot of business sense.
Though I imagine that if Haskell was given more of a chance in enterprise settings, the level of software quality would go up, though it may cost a bit more to write.
The Haskell community shouldn't try to do better than Java. Enterprise is what it is because it has particular motivations. We should play to the strengths of our language and our community.
It's easy to agree that there's lots of software floating around, but that not a lot of it is good. That's particularly true about enterprise software (I used to work in the field). It doesn't have to be good. You could have an excessively complicated UI and do your numerics in Python (for the love of God, numerics and statistics in Python?) and people will still use it. The damn app can take 12 seconds to load, and people will still use it. How are you going to sell Haskell to the enterprise community when that's their bar?
If people want to write code in Haskell for a living, if they're serious about this freeing-software-from-the-von-Neumann-paradigm thing, they should seek out problem spaces where software quality matters. This is hardly the sole core competency of Haskell, IMO. When one of my friends was introduced to it, he noted that it was a very "scientific" language. Which is to say, he thought it was impractical for regular software. The flip side to this is that we can very easily do things like check for gene sequences in Haskell (Brian O'Sullivan was doing something like this).
The fact that things like sets and maps are treated like values means that they are very natural and easy to work with (albeit a little slower than their mutable counterparts in other languages), so why aren't we seeing Haskell used to analyze social networks? Haskell is a powerful language. We just need to see it for what we could do with it, rather than redoing the old things that have already been done.
Though I imagine that if Haskell was given more of a chance in enterprise settings, the level of software quality would go up, though it may cost a bit more to write.