The Future of Automation

By Peter L DeHaan

January 9, 2008

In this issue of eConnections, I will expound upon "computer agents," the fifth of six observations that I gleaned from the 2007 ATA Convention & Expo.

Of the six items, this one has only a tangential connection to the convention.  To my recollection, no speaker directly addressed this topic, so it results more from the seemingly random comments of others and my own inferences, which were made possible by being in an environment that was conducive for insightful thought to occur.  Conventions are good for that, especially ATA's.

First, a definition is in order.  By "computer agent," I refer to the personification of a computer algorithm to imperceptibly mimic a specific, real-life human agent in sound, in words, and in manner.  Think of a computer agent as a personal telephonic avatar. 

As an example, my computer agent would sound like me, employ my vocabulary, and construct sentences just as I would.  It would listen to – and understand – my callers, able to correctly ascertain the reason for their call, and appropriately respond.  If my telephonic avatar wasn't able to properly or effectively assist the caller, it would make a seamless handoff to the real me, so I could take over and complete the communication.  The caller would never discern that they had first been talking to a computer-generated persona of me or that I took the call over midway through.

Of course, the computerized version of me would have at its avail all the information in my computer database – and the Internet.  Whereas I need to take a perceptible amount of time to determine if an article has been received or verify that an invoice had been paid, my computer agent could quickly access that information and effectively relay a cogent response to the caller in real time.  I could even program my computer agent to communicate more effectively than the real me, correcting those grammar issues that I struggle with, deleting the ums and ahs that creep into my speech, and even correctly pronouncing those words that I regularly state with imprecision or in error.  Best of all, my computer agent would consistently exemplify me when I am at my best.

In the call center, computer agents – on behalf of real-life counterparts – would answer all calls and handle routine information gathering and dissemination.  On harder transactions, the computer agent would pass calls to their real agent counterpart – along with relevant data, both obtained from the caller, as well as supplemented from other sources – for an appropriate call resolution.

At first, a computer agent might completely handle 20% of all calls, passing on the remaining 80% to its real agent counterpart.  That's a nice productivity boost, but it's just the beginning.  As the technology improves, that automation ratio will assuredly increase.  When it surpasses the 50% mark, it then becomes realistic to duplicate the computer agent, having two computer agents simultaneously working for each real agent.  Perhaps computer agent efficacy might even someday approach 100%.

Although technological progress is mightily advancing in this area, I am resigned to declare that the fully functional computer agent, as I have articulated herein, is not yet a reality.  However, it will most certainly come to fruition.  It is only a matter of time, be it two months, two years, or two decades – it will occur.  I can hardly wait.