It's easy to blame IVRs for a lot of your company's customer service problems. Because it's true-sometimes the IVR just isn't very well designed. A lot of times, though, it helps to look past the IVR into business practices.
Here's an opinion piece that takes issue with a Gethuman recommendation to send callers to a live CSR if one is available. The author of the article points out correctly that there's value to companies in identifying high value customers, and giving them preferred service. However, you can still prioritize customers without violating the Gethuman principle he identifies.
I have other problems with the Gethuman principle, and with the author's unstated assumption in the article. Both assume that IVR contacts are necessarily bad, or at least worse than a interaction with a human. For simple interactions that involve call transfers, account balances, password resets and so on, there's no reason why the IVR interaction should be less satisfying than one with a CSR. Better, in fact, if you add a little time in queue to reach a CSR. The fact that many IVRs give such poor service even for simple requests is due to failures of design and implementation, not to a limitation of the technology itself. You can serve simple requests from high value customers using IVR just as well as with CSRs if your IVR is well done.
The author identifies IVR as the culprit for poor customer service. In many cases, that's true, the IVR is poorly designed. The other, harder problem is that many companies are bad at identifying high value customers, and in defining what it means to provide high level service. What good does it do to get a caller to a CSR who mumbles, can't find information, doesn't recognize opportunities to provide additional service, doesn't answer the unasked question, and so on? Those problems aren't solved with even the best IVR.
The upshot of this is that IVR designers need to be able to see the big picture, and understand what problems the business is trying to solve. If they can't do that, they'll wind up spending a lot of time designing IVRs that don't address real issues.