When it comes to Enterprise Applications which is commercially dominated by SAP and Oracle, there are several open source solutions being developed. Some of them are trying to be the complete suite of applications like Compiere and the others more best-of-breed point solutions like Coupa.
Just like the CIOs and IT staff face with the dilemma of picking the best-of-breed commercial solutions vs going with generic but wide range solutions, same is applicable for open source as well.
Moving from paper based tracking to any form of automated tracking certainly helps. But for companies trying to go beyond the basics and have well integrated automations from procurement to payment, order (or even a quote) capture to fulfillment, there will be integration challenges to deal with if one were to go with best-of-breed solutions. Infact, one can even imagine a configuration where some of the best-of-breed solutions are commercial while the others are open-source!
In such a scenario, who will be building the integrations for these disparate best-of-breed open source applications? Would it make sense to bring in the big-5/4 there? IMHO, then that would defeat the purpose of trying to use open source with the intention of saving the bottom-line.
In addition, with open-source applications, there is no restriction on the technology components. One team/company can choose to use Ruby-on-Rails, while the other perhaps Python and yet another Java and may be even Perl. So, what a nightmare that would be to integrate these various different technology components into one single harmonious application. Of course, this is true for commercial counter parts as well. However, most commercial software typically goes with the main-stream technology, which currently is either Java or .Net.
There is no denial that the WebServices and the SOA are supposed to address this language/platform problems. However, it is yet to be seen how successful this route is going to be as many earlier attempts at seamless distributed computing (such as CORBA and RMI) never really worked well mainly due to performance. But even with WebServices and SOA, most of the open source enterprise applications are currently at a stage where they need to catchup and implement a lot of functionality and the least focus they have is to ensuring that their application is interoperable with other applications. That is not to say that they are not interested in that, but with limited resources, there is only so much that can be done and that would typically be on the core competency, the product itself and how many features it has as a standalone application.
But who knows, with the commercial applications already betting and leading their way to SOA enabled applications, the open-source counterparts will consider it as a part of the survival strategy than a add-on bolt. I see the same issue with business intelligence also where the main stream commercial applications have some in-built capabilities while the open source counterparts mostly lack them.