Software whose failure everyone notices quickly.
Probably a good definition, but not for the reasons you'd think. If you've ever been involved in a Java/COM/.NET "enterprise" adventure, you'll know that "everyone" means the CEO/CIO who's gotta fire somebody, and quick!
My definition of enterprise software ?
Something that costs at least one order of magnitude more than a reasonable solution to the problem should.
Before you dismiss my definition, consider this case in point:
I've developed both a Perl DBD, and a JDBC driver, for a large scale "enterprise" database system. When I attempt to solicit a modest price for the DBD, I am greeted with silence and occasionally derision. When I ask >US$10K per license for the JDBC driver, many sites happily fork over the money and the JDBC driver was derived directly from the Perl driver!.
I suspect most of us know that many (most?) things being done under the auspice of "enterprise" J2EE could easily be delivered with Perl (or python, ruby, PHP, et al) for a 1/10th the time and money. But Perl/Python/Ruby/etc. don't have an army of marketing suits telling PHB's that its "enterprise".
In summary, "enterprise software" eq "marketing grot".
If your purpose is to market Perl, then you'll need to discuss its scalable paradigm shift in the context of re-engineering the enterprise to repurpose the legacy platform solutions in the context of grid computing frameworks for exposing service oriented architecture access via the empowering technology of dynamic language implementations.