Here's the nub of the problem with conventional business software: "Companies are paying too much for complex products that they can't fully control." One such company, described in this article, chose to solve this problem by bowing out of its existing contract for Salesforce.com's on-demand CRM service and replacing it with open-source software from SugarCRM:
"[Athena Healthcare's CTO Bob Gatewood] went open source, teaming up with the Cupertino, Calif.-based start-up to design and then graft his own CRM system onto his existing IT infrastructure, helping his company's bottom line and boosting employee efficiency across converging software platforms."
I wonder, though, how much control you get in the long run from tightly integrating your systems to a specific software implementation? SugarCRM's business model is founded on users of its software coming back to it for support and customization services, and despite its open-source credentials, the company keeps a tight rein on product development (that's pretty much the default practice among open source vendors, as Mambo's developers have recently discovered). So even though you could, in theory, cut free from SugarCRM and pursue your own development with full access to the source code, it's hardly likely to be an economically viable proposition. The reason packaged software exists is because of the economies of scale that result when vendors can spread the cost of product development across many customers.
That's why the (progressively more lamentably named) sforce, Customforce and Multiforce projects at Salesforce.com are such critical components of the vendor's forward strategy. Their purpose is to deliver control to customers using a more flexible, loosely coupled services model than the conventional model of software integration that Athena Healthcare has adopted with its SugarCRM project. If on-demand providers can combine lower cost with greater flexibility and control, then they may yet trump conventional application software, even when delivered under open-source licensing.