How Chatbots Can Create a New Kind of Agent

‘More human than human’

—White Zombie (‘More Human Than Human’)

In my day-to-day job, I spend a lot of time talking to companies about their customer service strategies and technologies. And pretty much every day, companies want to talk to me about conversational artificial intelligence (AI). It has gotten so extreme that I often joke that chatbots ate my life. (Hey, I’ve warned you, dear reader, that I am no comedian, no Chris Rock, no John Mulaney. Heck, I’m not even Gallagher or Carrot Top.) Undergirding all of these discussions is the idea that companies will reduce their dependence on their human agents, either by being able to handle more interactions without adding new staff or by actually reducing contact center head count.

But as you might have guessed, there is a twist…and hopefully more like a Jordan Peele twist than an M. Night Shyamalan lame-o twist. The twist is this: The more that automation succeeds, the more that humans matter in customer service—they just matter in new ways.

Let’s start with the assumption that at some point the hype around conversational AI comes true and chatbots actually work at providing us relevant, contextual, and personalized answers and also start to execute transactions at our behest. In such a future, what we think of as tier-one agents start to become unnecessary. But this automation will change the remaining customer service jobs: If we don’t have tier one, we still need folks to provide human empathy, to solve truly complex problems, or to use human intuition for corner cases. In short, we will have tier-two or tier-three contact center jobs. Those are more highly skilled and more highly paid positions and more resemble what we think of as knowledge workers today.

Which is all great—well, sort of. We’ll need fewer of those people, so likely fewer jobs. But to make up for it, it isn’t too difficult to envision new types of jobs that automation creates. Maybe we take those excess contact center agents and have them act as what I like to think of as “bot wranglers.” Just picture the cute rodeo-inspired outfits those workers could be rocking in the office! A bot wrangler would oversee multiple chatbot conversations, observing the interactions and injecting human expertise when required.

Here’s an analogy to explain the difference between a bot wrangler and the current model where humans take over when chatbots fail customers. Let’s use toy trains as our analogy. In today’s approach, if a toy train goes off the track, we have a human being pick it up off the table, place it back on the track, and then manually push the train around the track until we are done playing for the day. Inefficient, obviously. And maybe fun for the person pushing the train around, at least for a few minutes until it becomes tedious. In the bot wrangler world, the human again picks the train up off the table and places it back on the tracks, but then the human pats the train on the caboose and sends it on its merry way. Elegant and efficient, as befits the rhinestone-studded bot wrangler getup.

A bot wrangler could be helpful in interpreting some regionalism for which the bot wasn’t trained. For example, understanding the difference between youse, y’all, and yinz. (Wait—is there a difference between those?) Maybe a more common function: The bot wrangler would disambiguate customer intents. If a customer says, “My credit card is toast,” does that mean “Replace my credit card,” or “My credit card was stolen,” or “An ATM retained my credit card”?

So, go ahead and dream of an all-machine future; I’ll be trying to design cute matching bolo ties for the bot wrangler uniforms of the future.  

Ian Jacobs is a principal analyst at Forrester Research.