It seems like everyone is getting into the ChatGPT craze lately, and customer service teams are no exception. So what’s the right way to incorporate AI (artificial intelligence) and automation into the customer lifecycle?
I recently had the chance to talk with James Diel, Chief Business Development Officer at Capacity, for a new podcast called Practical AI: The Capacity for Good. Below is an edited transcript of a portion of our discussion that covered ChatGPT and its use in customer service.
James: Let’s talk about the thing that is top of mind for so many people right now. AI automation: ChatGPT. So based on your experience, what is the impact of AI automated customer support on the experience? And how can that help companies gain a competitive edge?
Related: How to Use ChatGPT to Enhance Your Customer Experience
Dan: So first thing I think is we’ve got to have the right philosophy. Every time there is new technology, whatever it is, there are some companies that say, ‘awesome, this is going to be great because now we don’t need X again.’ And in the case of, say, chatbots, it was ‘we could get rid of customer service people and save millions of dollars.’ That is not the philosophy that I would advise.
Technology is wonderful, and it’s getting better and better. I am marveling at the fact that we are still in the early stages of ChatGPT. We are still at the very, very beginning of all of this. And it’s mind blowing, right? So it only gets better over time.
But let’s not look at it as a replacement. Let’s look at it as a supplement, as a complement.
I always like to imagine, for example, remember IBM’s Watson, the one that went on Jeopardy! and beat all the humans? Well, I always love to imagine in my head a customer service agent that’s sitting next to Watson. So let’s think about this agent. This agent has all the confidence in the world because they know that any answer to any question, they’ve got it right here. They got the Jeopardy! champion right next to them, right? So they know they’re going to know the answer to a question.
They also know that they’re not going to have to spend time looking at 17 screens and opening all these windows and typing this and clack, clack, clack. Well, if we now put ourselves into that customer service agent’s body, our job now is just to be human. It’s just to have a conversation with the customer.
James: To make that human connection.
Dan: To make the connection and have a relationship, right? Because we’ve got all the answers. Now, that is a fabulous use of AI, because it’s indifferent from saying, ‘oh, we don’t even need Sally the agent, we can just have Watson do it.’ Well, there we’ve lost human interaction. And so to me that’s not as good of an answer.
But when we have a fully informed, fully confident and focused agent, Sally, who is now like ‘my whole job is just to relate to the customer.’ now we’ve got a great experience for the customer because we have a nice person on the other line who is helpful and friendly.
James: They’ve got all the answers.
Dan: So now that’s to me the perfect scenario. Someday maybe the technology will be good enough to look and sound like a human and have all of the answers and finally replace Sally’s job. But I personally think we’re still quite a ways from that.
James: Yeah, I tend to agree with you on that. I think especially that there are so many things that customers today feel like a chat conversation with a bot is sufficient for. But you get into a complex situation, you want some empathy, you want some understanding that the AI just doesn’t have today. You have to talk to a human. That escalation has to be seamless and it has to be quick in order for that experience to even work.
So what tactics can or should companies use to measure the success of their customer support automation and its impact on customer service? Do the metrics change? Should how we measure success in that space start to look different now that we add automation and AI?
Dan: I always like connecting metrics as quickly as possible to the bottom line because at the end of the day, when you’re talking to a C-level executive, that’s all they care about.
Let’s just take this from a CX perspective. There’s lots of CX teams, and they talk about their NPS score, Net Promoter Score. It’s a very good indicator of how we’re doing at a moment in time, but it doesn’t tell us why. And so I have witnessed this time and time again: when it goes up, everybody cheers and we ring the bell and we go to the Slack channel and we say, ‘we’re so great, our NPS score went up.’ And when it goes down, we say things like, ‘oh, it must be the pandemic or the weather or climate change’ or whatever. We blame it on something. The truth is, we have no idea either way. None. And so what we have to start doing is tying metrics like that back to the bottom line.
Related: Customer Experience Metrics: The Top 5 Ranked and Could NPS Detractors Be A Good Thing? New Evidence Says Yes
So let’s take your example. The first thing I would look at is some metric around the accuracy of the AI. And that’s a pretty objective metric, right? How many times out of 100 does it give an answer that the customer is happy with? And even with ChatGPT, because it’s a conversation you can measure. How many times do I have to clarify myself until you actually give me the answer that I want?
Quick example: the other day I was writing a blog about the new Starbucks CEO and how the first thing he did was get his barista certification. And he also committed to spending a half a day a month in a store as a worker. And so I asked ChatGPT what other CEOs have worked in their stores? And it gave me an answer, but it wasn’t what I was looking for. The answer it gave me was, here are 10 CEOs that started off on the front lines and rose up in the ranks to become a CEO. A good answer to some question, but not the question I was asking.
So I had to clarify it. I wrote: How many CEOs are working on the front lines while they’re CEOs? And it gave me a different answer. So we can objectively measure the percentage of time that the computer is right versus not right. Now let’s link that to some satisfaction measure because we’re going to have a one-question survey at the end of each engagement that asks the customer, ‘how satisfied are you?’ Or maybe it’s an NPS question about ‘would you recommend us to a friend,’ et cetera.
And we can see that all the times that it’s right, our NPS score is here, and all the times it’s wrong, our NPS score is there. Now, we take NPS and we link it to the retention of that customer, the retention rate, how long does that customer stay our customer? I think customer retention rate is probably the most important metric that any B2B should be measuring. Because, again, as I said before, if 100% of the customers we bring in the door never leave, our entire business looks a whole lot different than it looks today.
And so if we can take the accuracy of the AI, connect it to the satisfaction, and connect the satisfaction to the retention rate, now we can tell our executives the accuracy is driving retention. And that is a number that they understand because they know the value of every customer is X. So the more customers we retain, the more money we make.
James: Great structure for being able to look at that. And those are the kinds of things that I think businesses are going to have to rethink, is how do we look at what AI is doing for us or not doing for us and how do we, to your point, draw that to the bottom line.
Dan: Yeah, and I suppose there is a piece on this bottom line thing, to be fair, where there’s probably a cost savings as well because the AI probably has replaced some hours of human work. And obviously in your calculation, you’d add that in as well.
James: Yeah, absolutely. So I’d like you to put your kind of futuristic hat on, think about AI and customer service in the future. So as this evolves, what do you think customer support looks like a decade from now? And maybe even it’s two years from now. What are some key things you see changing or expanding from what we see that today?
Dan: I think that one of the things we often either don’t talk about when we talk about AI or we mush it together like it’s the same thing, is machine learning. Machine learning is what makes AI smarter. And so as the machine continues to learn, it gets better and better. And we’ve seen this just through a few versions of ChatGPT. It is going to keep getting better and better.
I think as it does, consumers are going to trust it more. There’s a lot of skepticism right now, but as it gets better, that skepticism will go away. And I do think there is a time, I don’t think it’s right now because humans, especially post pandemic, crave human interaction. But I do think over time, as this gets better and we start to trust that, yeah, I can ask this robot pretty much anything, and I’m going to get the right answer, we’re going to feel better about it.
And I do think, whether it’s 10 years down the road or whatever it is, that we will start to see the vast majority of customer service interactions happen without a human. I still think you will always have those super technical, super complicated, once in a lifetime, unique questions that you’re going to need a human for. But I see that percentage going way down as machine learning goes way down.
James: Yeah, makes sense. I mean, I think right now there’s nothing more frustrating than a bad experience with a chat bot. But I also think a great experience is one where you say what you need, they understand it, they get your answer and you move on. That’s an amazing experience. So when it’s good, it’s great, when it’s bad, it’s horrible. Is there anything else that we need to think about the B2B customer experience?
Dan: Yeah, anytime you’re in a meeting and the conversation goes to a place where somebody says, well, we’ve always done it that way, or we’re doing it this way because all our competitors do it this way, those should be bright, loud sirens in your head saying, this is not how we should do it. Because if all of our competitors are doing it this way, then we’re not going to be differentiated. And if we’ve always done it this way, it probably means it’s old thinking, right?
James: Great insight. Very practical and applicable to businesses and B2B and what we’re dealing with right now, trying to identify how we differentiate ourselves in an extremely competitive market customer experience.
Image by kiquebg from Pixabay. Video courtesy of IBM Research.
Listen to the full interview on the Practical AI: The Capacity for Good podcast.
