The Art and Business of Speech Recognition: Creating the Noble Voice

Before we discuss the benefits of deploying in multiple phases, let's examine the pitfalls of not deploying in phases.

Let's say we're operating a large call center. In anticipation of launching our new speech-recognition system, we've rerouted our phone lines, and trained and reassigned thousands of agents to handle other tasks . The big day comes, we throw the metaphorical switch, and ” ta-dah! ”our system is fully deployed.

But there's a problem. Perhaps the system can't handle the call volume for technical reasons. Or maybe we start getting hundreds of calls from frustrated or angry callers who were given the wrong information by the system or who stumble over a particular question (that was never encountered by usability testers) almost every time. We'd probably have to shut the system down to fix these problems. That means our agents would have to return to their old responsibilities ”at least temporarily. Our phone lines might have to be reconfigured. And in all the pandemonium and confusion, we might be completely offline and unavailable for hours. Imagine the cost ”in dollars, in customer dissatisfaction, in employee disgruntlement, and potentially , in bad press. It's simply not worth it.

The insights and peace of mind we gain by deploying a system in phases greatly outweighs any expense in resources or time associated with it. In fact, to have a successful transition to a speech-recognition system from a touchtone system or from live agents, a planned deployment is a requirement.