Plain and simple: the bottleneck if you are dependent on software as a tool will always be the software; and the bottleneck for advancing the state of the art for software will always be software developers. If you are not a programmer and you use software to solve problems at work, you will always be beholden to this simple and often annoying arrangement.
From this, you have a three choices: do it yourself, pay someone else to do it, or hope somebody does it and shares it with you. If you're not a developer, the first option is out. If you can afford to hire a programmer and/or license software that already does it, you can go with the second option. If your workflow is commoditized and simple enough that it's in the open source stack (and more and more gets to this point all the time) then you can use OSS, with the awareness that option #1 still stays in play: you're dependent on software developers who are either scratching an itch and have no compelling reason to support you all the way through your journey of discovery, or you are paying large license fees to have throats to choke for software that is delivered as a black box but fulfills your needs.
All in all, this is a complicated issue. Software to this day is still hard to write and still requires programmers of some level of skill. Programmers like paying their bills. Whether you find a way to leverage the work of others through common interest (off-the-shelf open source, grant-funded projects) or by paying for work (sponsoring developers to contribute to open source, getting grants for vested third parties or buying commercial) is a per-case judgement call. I think in general buying off the shelf is still cheaper than hiring programmers to help out because the cost-per-person is amortized across every organization that pays a license, but you might get better domain specific tools by hiring people to work for you locally. And if you have enough smart people already working for you, the cost of using/getting productive with open source versus the time spent learning its quirks will look fiscally viable.
The comments on this recent thread should also prove helpful.