I see real value in real time analytics for the contact center, particularly in terms of its ability to gather valuable customer information which can be used across the enterprise. But as a former outbound agent from the 1980s, I’m distressed about this idea of using real time speech analytics to identify calls that are going sour and automatically intervening on the call for the purpose of preventing the customer from defecting. From my recent discussions with company officials in the speech analytics field, it sounds like speech technology has not advanced to the point where you can rely on it for this purpose. Maybe that’s why real time analytics continues to be referred to as the “Holy Grail of the contact center industry.”

Don’t get me wrong, I’m not against automation, especially when it leads to operational efficiencies and higher customer satisfaction. I think it’s fine that a contact center supervisor can monitor a potentially failing interaction and immediately send a screen pop or instant message to the agent telling them, in real time, how to calm down the customer so that they don’t leave. I also think it would be fine to use real time speech analytics to automatically identify calls that are in trouble, through word or phrase-spotting or “emotion detection,” and then using that to automatically trigger screen prompts telling the agent how to respond. My problem is with intervening on the call — especially if you use speech analytics to make it happen automatically. To me this sounds like a case of over-engineering, and I can imagine it causing serious problems for agents. Furthermore, I see this idea as being in conflict with the current movement in the contact center industry to empower the agent.

First, I think it can lead to more confusion, since very often the intervening supervisor isn’t aware of the full context of the call, since he probably has not listened to all of it. Second, it can be a humiliating experience for the agent, especially when the supervisor (or “automated supervisor”) commandeers the interaction. Personally, I think its better to train your agents well before they get on the floor, and then use “silent” systems, such as screen pops, scorecards, dashboards and e-learning, to deliver ongoing training and coaching in real time - or near real time. I’m not sure I like the idea of intervening via “whisper coaching” either. I can just imagine what it would be like to be on a call with a customer and have a supervisor come onto the line and whisper “cross-sell!” into the headset, as I attempt to navigate the customer through the company Web site. Although I’ve never experienced it, I could see this as being potentially disruptive — possibly causing the interaction to take a turn for the worse. I can also imagine a speech analytics solution misinterpreting what is being said and automatically delivering the wrong prompt, thus throwing me off, causing confusion.

Furthermore, I don’t see this as impressing the customer, particularly if he or she is aware that someone, or something, has intervened on the call. Doesn’t that reflect poorly on a company’s training practices? I also fear that these automated systems might cause companies to skimp on initial training and rely on the automated systems for performance improvement after the agents get on the floor. That means using your customers as guinea pigs.

Anecdotal evidence suggests that such automated systems, when used to improve agent performance, can also have the effect of negatively shaping agent behavior. My question is, if you’re going to bring the technology to this level of sophistication, then why not just automate the whole process using speech recognition? With fully-configurable, fully-automated, naturally-conversing self-service speech solutions now available, I wonder if it’s worth it to work so hard to try to save every customer and correct every little bit of human error in real time …

The author can be reached at [email protected]