Five years ago, when you called your credit card company or any other customer service you had a choice: either punch in your selections in the automated menu or hit the O-button to get to a live operator. Back then everyone hated those annoying automated menus. Little did they know what was going to happen in the future thanks to technological advances…
Today, most of the customer service departments use some voice recognition software, which supposed to collect all required information (e.g. your credit card info, address, the reason of call et al) before the live customer service representative would start talking to you. The goal is noble, but implementation sucks.
After spending 5 minutes repeating like a parrot the same answer several times, you may not be understood anyway. The old trick with keying in the O-button doesn’t work any longer and just pushes you back to this cold-blooded machine, which keeps asking you politely, “Did you mean to say…”
Finally, it understood what you wanted to talk about, validated your credit card number, spent another minute announcing your balance on the account, which you never asked for, and when you’re almost ready to throw your phone against the wall, a pleasant voice of a human being says, “Hello, this is Adriana. For security reason, could you please tell me your credit card number?”
WTF, didn’t I go through all this already talking to your stupid machine?
I’m sure, the CEO of these companies receive good looking reports stating that “our company was able to substantially minimize the time spent by our agents on the phone with customers, which allowed us to lay off 100 employees resulting in $5M of annual savings.” These reports don’t account for losing business because some customers just hated this type of phone experience and went to competitors that operated in an old fashioned way with live service representatives. Mediocrity an incompetence in enterprises disappoints. But there’s not much we can do about it other than learn how to live with it.
yakovfain.com 
Your frustration with call center implementations of speech recognition is clear and warranted in many cases. To the degree you are blaming the speech technology, however, you are missing the mark. The decision on how to structure the automation is made by call center managers using what I have called “touch-tone mentality,” that is mimicking what they used in a touch-tone system and not letting the speech recognition make the process easier, certainly easier than waiting 5 minutes for an agent that may or may not be well trained. The natural-language speech recognition in Siri gets one to an answer or a service, usually in one step. It turns out that this technology is available to call centers and used by some. The first prompt in some of those systems is simply, “How can I help you?” One company offering this technology is Nuance Communications, which supplies the speech recognition in Siri. And Nuance is teaming with IBM to use IBM’s Watson knowledge representation technology to get you an answer even more quickly–they call the new service Prodigy.
The problem is poor implementation of this technology as I stated in the blog. They forcefully turn customers into early adopters of an immature technology. The solution is simple – start any automated voice recognition system with a message, “If you prefer to talk to a live operator press 0 at any time”. As simple as that. Just give your customers a choice till Siri or any other technology will come really close to being a decent AI solution. But even then, I want to have a choice of talking to a person.