Today’s Speech-Enabled Self-Serve Solutions Stand to Rob Speech Analytics of its Efficacy

Patrick Barnard
Group Managing Editor, TMCnet

Today’s Speech-Enabled Self-Serve Solutions Stand to Rob Speech Analytics of its Efficacy

As I’m sure you already know, speech recognition and speech analytics are two related yet very different disciplines. In the call center world, speech recognition has evolved from the clunky DTMF touchtone solutions of a few years ago (which you basically bolted onto your switch to handle call steering), to highly advanced, speech-enabled self-serve solutions -- or “automated agents” -- which actually have “personas” and can guide you, just like a live agent, through a transaction via natural conversation (this facilitated through today’s advanced speech algorithms and increasingly sophisticated “full vocabulary” speech engines). With today’s self-serve solutions, very often the “automated agent” can handle the entire interaction – whether it’s informational or transactional -- and the customer may never need to speak to a live agent. Speech analytics, on the other hand, primarily relates to the “mining” of recorded call data, which typically takes place post transaction, and does not require the same level of speech recognition technology as an “automated agent” or IVR application takes. With speech analytics, organizations can “mine” their recorded agent/customer interactions, post-transaction, and gain new insights into customer and agent behavior – insights which can be used to drive key business decisions.

OK so there you have the basic functions/advantages of both technologies. And keep in mind that both are now largely viewed as being an application residing within the enterprise network – not just on the call center switch. I won’t bore you with my detailed list of the advantages of each – nor will I bother explaining how the value both of these technologies becomes exponentially greater when they work as part of a tightly integrated suite of applications. Suffice it to say that both are now being viewed as important, if not critical pieces of the contact center software ecosystem – and speech analytics is now being viewed as a critical tool for gaining business intelligence.

Here’s the point I want to make – and I make this point every time I do an interview with a company that deals in both technologies:

Speech recognition will one day begin to rob speech analytics of its value.

Why do I say this? Well, it’s a pretty simple observation, really. As more and more contact centers begin to deploy speech-enabled self-service solutions -- and as they are able to have those systems handle more and more of the transactions all on their own -- there will obviously be less need for customers to interact with live agents. But wait a minute – what about those important “insights” that are delivered through the mining of recorded interactions? What happens if a large percentage of your customers never actually talk to a live agent?

You see, as call centers push off more and more calls onto these automated systems, they are losing the valuable customer insights that they normally gather by mining recorded interactions through speech analytics. It’s kind of ironic, in a way, but one technology ends up sapping the power from the other. If you have less recorded interactions, then you have less data to mine, right?

Usually when I raise this point, most vendors explain that very few organizations are able to serve their customers entirely through self service solutions – in other words, there will always be an opportunity for a customer to talk to live agent when there is a problem. But with the industry really pushing for these self-service solutions, and with these solutions becoming more and more advanced each day, it is easy to imagine that one day most customer service centers will be able to handle the vast majority of calls using self-service apps. So, my point is, even if a call center is able to automate 60 percent of its transactions – that’s still 60 percent less recorded interactions which can be mined to gain valuable business insights. Now you can only mine 40 percent of your interactions – and those are probably the interactions where the customer was dropped out of the self serve system for one reason or another. In my view, this totally stands to “skew” the data that is collected from through speech analytics. Now, you are able to mine only the transactions where something went wrong, initially, or at least only the ones where the customer had to speak with a live agent (perhaps the call was escalated to a higher level to handle a higher value transaction, or whatever).

I see tremendous irony in this and I’m wondering if there are any other readers out there who also see that speech enabled self-service solutions are like a “knife in the back” of speech analytics. I’ve been told that some organizations are now recording and mining the interactions with the self-serve solution, to see if they can gain insights from those, but to be honest, I don’t really see the point in doing that. I mean, after all, if you’re customer completes his transaction with the automated system, you must be doing something right … right?

Anyone out there think I'm right?

Leave a comment