The successful development of voice services capable of replacing humans in call centers could save businesses billions of dollars. JLL Research estimates that globally call centers are costing $310 billion per year.
The race is on to develop the smartest and most charming voice AIs, with leading tech companies like Amazon, Google, IBM and Cisco all investing millions into the research and development of voice technology.
With millions of dollars behind it, the scope of voice applications is expected to grow fast. I started to wonder: Coud a robot ever be as effective as us mere mortals in a customer service role?
What happens in call centers?
When I was 17, I was let go from my seasonal job at a call center.
The call center belonged to a popular TV shopping channel, and my job was simple: Listen to people’s orders and put them into the system.
The reason I wasn’t among the cohort that got kept on, despite (at least I thought) being very friendly and highly resourceful, was that I had, during my time there, been the unlikely recipient of a formal warning.
The reason I got a formal warning was that my call time had been way above the target average call time of 90 seconds.
The reason my average call time had been way above 90 seconds was that at around 7 a.m. I’d accidentally talked to an elderly lady for 90 minutes, instead of the recommended 90 seconds. I suppose I was secretly a human being playing the role of a robot.
Our remit was incredibly limited.
We were authorized to perform around three very standardized functions. We could change someone’s password (but, of course, never give out the old one), update bank details or simply place an order. All of these things could be done in under two minutes, and there was a set script for all of them.
We spent the day repeating ourselves over and over again, like robots, but trying to keep as much human energy in our voice as possible. We struggled to scribble down people’s preferred name on paper, so that we could surprise them by still knowing it at the end of the 90-second call.
We weren’t there to be human — with hundreds of employees it’s rare to speak to the same customer twice. Our goal was to become the most efficient version of ourselves, staying on script, and competing against our own and others’ average call times.
A robot could do this.
In many regards, yes, a robot could have done it. But, the reason that up until now this job has been reserved for humans is this: While the call center representative is generally highly scripted and predictable, the caller is not.
Human communication is idiosyncratic — we have different pace, intonation, accents. The human brain has learned to quickly adapt and understand others. When you speak to a new person, your brain quickly figures out which accent they have so that you can tune in effectively.
For example, you could hear “I slipped bedly because of my bed bid,” but then once your brain has recognized the New Zealand accent, you realize the person actually said, “I slept badly because of my bad bed.”
Until recently, machines have really struggled to understand the nuances of human communication. As humans, we rely a lot on context to help us decide what somebody is saying. Now, machines are developing the artificial intelligence to be able to understand millions of variations on human speech, and also use context and machine learning from a wealth of previous experiences to find the right answer.
As machines learn to understand our every utterance, the exciting task will be training them on how to respond. Over the next decade, user experience designers will have their work cut out designing voice services that are pleasing to interact with, do not frustrate the user and are as efficient as a human call center representative.
Someday all call center humans could be fired.
In the near future, the AI behind voice services probably will mean that humans in very repetitive customer service roles will be liberated — and will have to find new ways to earn a living (perhaps by bicycle-delivering fresh healthy food to the experts building voice AIs?).
And why did I always spend too long on the phone? Yes, I can babble on, but there are also some situations that are difficult to speed up. If someone on the phone can’t remember her password, she often looks around the house for ages, insisting, “I must have written it down somewhere,” and asking various relatives about what or where the password could be.
Maybe a robot would have less problem simply saying, “Just take your time. Don’t you worry, I’ll make sure you get put straight through to me (mwahaha, we’re all me) when you call back. Goodbye.”
When I was leaving, my manager confessed to me that she’d also wanted me to leave call center life before I got sucked in. I’d mentioned taking a year out before university to work there full-time and save up — her own plan many years before. Maybe she knew I’d have still been there 20 years later when the pink slips were handed out and the robots moved in.