With the astonishing advances in AI in the last three years, it might be a good time to reassess the importance of the human touch in customer service. I am amazed at what the latest wave of generative AI can do, on top of the tremendous advances we’ve been living with for a while already. A new horizon is coming into view of infinite ways to apply these new technologies. People are discovering ways to combine them with other technologies that already exist or are now being developed. We are only at the beginning of learning what works and what doesn’t. Nearly any repetitive task can be automated. But some customer relations functions are better performed by people.
All computer tasks are part of larger systems that include human input at some point. There is always a point at which the computer’s work is done and the job passes back to the human domain. In the realm of customer service, it’s better to discover that point early, before you unnecessarily drive away customers mad.
It’s a time for experimentation, to discover in practical terms how these potentialities can be applied, what a system can do, and what it can’t do.
Electrifying!
I liken this moment to the introduction of electricity into human civilization. When Thomas Edison first wired New York for electricity, he was thinking about lighting. That itself was an enormously ambitious, world-changing achievement. But making electrical energy generally available created infinite possibilities, and inventors got to work figuring out ways to use it. It led to the world we now live in when our homes (what Buckminster Fuller called our “living machines”) are primarily driven by electricity. Electricity is part of the functioning of practically everything we use. When the electrical grid goes down during a storm, we learn quickly how dependent we are on electricity for practically everything we do.
Now we are at a similar place in the development of AI tools, at the onset of a flood of innovation as these various new applications unfold, and the new generative AI meshes with previous technologies, from mechanical, to electrical, to electronic to cybernetic.
The strengths and weaknesses are revealed by trial and error. When it comes to using chatbots or interactive voice response systems in customer service, it’s important to discover problems and respond quickly, because the cost of losing a customer is high. Sometimes a little finessing can smooth over a problem, but it usually requires a human to make the right connections and suppositions to understand the human problem.
First, Do No Harm
People who come for customer service are usually experiencing some problem, and may already be feeling tense over it. Besides actually solving the problem itself, there is a need for a de-escalating experience with a sympathetic human being. Speaking to a voicebot designed to screen out callers and reduce the need for human involvement is often just the thing to set off rage in an already-frustrated person.
Just this morning, in negotiating entrance to a website I was asked for my birthday.
“September 21st.”
“Did you say October 1st?”
“NO! No, I said September 21st?” I exaggerated my enunciation.
“Did you say October 1st?”
“NO! That’s not it. I need to speak to a person! A representative! Please! I beg you!”
Talk about deaf ears. This is what you sometimes go through just to start a conversation. You know the drill. We’ve all had those kinds of frustrating experiences with voicemail systems. When you reach that point, it’s like putting your fist into a wall.
As we integrate AI into business practices, we’ll need to decide where to draw the line between what can be automated and what requires human attention. I think when it comes to customer relations, the ideal point for human intervention is likely to be pretty early in the process.
Stymied
A couple of days ago I had an internet outage. Knowing how these kinds of issues can eat up hours of time, I called Optimum, the provider, right away and settled in for a long, tedious process. The interactive voice response system rattled off some routine stuff, then asked me for my access code. I have no recollection of ever having one. The voice told me I can get the code by going to the website. Oh, thanks! But I had already told it that I’m calling about an internet outage, so obviously, there’s a problem with telling me to go to their site. And at this stage, I’m just trying to get started.
It’s annoying that the system is blocking me, telling me stupid things, but it makes it worse that it is completely mechanical and repetitive and has only certain possible responses to each thing you say. You can’t make a human appeal. It would be useless to explain why it makes no sense to tell me to go onto the internet to retrieve a code when the reason I called is because I don’t have internet.
Those are the kinds of customer service experiences that drive customers to the competition. In any case, voicemail decision trees are not the best way to perform customer service. Up to a point, the interactive voice response systems can be good time-saving devices. But at certain points, they can quickly cross over into obnoxiousness. By definition, good customer service is providing effective assistance before customers become extremely frustrated.
A Job for a Human
While I am blown away by what AI can do, I am also amused to discover the odd ways it sometimes falls short. The extra fingers on some of the AI-generated pictures are an example, and there are many subtle ways anyone, even a computer algorithm, can misinterpret or mis-contextualize a statement.
At the year’s end, Spotify gives its customers Spotify Wrapped which includes a year-end review of their listening habits. As an AI achievement, it was impressive.
My listening stats were used to generate a commentary delivered by AI-generated voices that sounded like a cute young woman and a cheerful young man, talking to me personally, teasing a little about what I listened to. Presumably, they did that for Spotify’s 250 million members. What they accomplished was brilliant. But it was also interesting how it fell short.
My AI avatars said I had listened to one song so much more than any others that I must be obsessed with it. But when they named the song, I didn’t recognize it.
Then I figured out why. It was the first track on a playlist that I play often because it has a Pavlovian calming effect on my dog. She’s an excitable Australian Shepherd, and I have discovered that this playlist calms her down and makes her relax when nothing else will. So, when I need to get some work done, I put that on and it’s remarkable how effective it is.
I put it on for the mood it creates, not for any one song. It creates a calm tapestry of soft guitar music. I put it on shuffle to keep it fresh, but I usually start at the beginning. So the first song got played every time I put on the playlist. The rest of the tracks in a 10-hour playlist only got played now and then.
The AI personalities made jokes about how obsessed I must be with that song and, by the statistics, it did look like that. But it wasn’t that at all. It was something completely out of the frame of reference of these devices.
The misinterpretation was harmless, but it was an example of how things can go off in subtle ways that can’t easily be foreseen, and why AI in general requires some human monitoring at some stages.
Those are the kinds of things a person understands intuitively, but even the most advanced AI systems may not understand. That’s the kind of job that requires a human. Fortunately, for the moment, there are still many of those around.
David Cogswell is a freelance writer working remotely, from wherever he is at the moment. Born at the dead center of the United States during the last century, he has been incessantly moving and exploring for decades. His articles have appeared in the Chicago Tribune, the Los Angeles Times, Fortune, Fox News, Luxury Travel Magazine, Travel Weekly, Travel Market Report, Travel Agent Magazine, TravelPulse.com, Quirkycruise.com, and other publications. He is the author of four books and a contributor to several others. He was last seen somewhere in the Northeast US.
