A month ago, I attended a conference hosted by a well-known provider of Artificial intelligence (AI) based software for contact centers. I was part of a small cadre of analysts who were given special access to their executive team along with some special demonstrations of current and forthcoming features of their systems.
AI and Machine Learning (ML) hold immense promise for the customer experience (CX) discipline. Real-time coaching is important in contact centers, but supervisors and managers don’t have enough time to do it widely and well. Now, AI can “listen in” on text or voice conversations with customers and provide suggestions to agents in real time, either on-screen or in-ear. These coaching systems can fill in the blanks for agents about products and about options to make up for less-than-desired experiences, taking anxiety and delay out of the conversation. These tools can also be used to review the thousands of hours of recorded conversations that would otherwise languish as “dark” (unreviewed, unused) data in digital storage, performing analysis that humans simply cannot do.
Another front on which there has been significant progress is Conversational AI, which uses Natural Language Processing (NLP) to converse in human-like ways. One company (not the same one as the conference host) has even set up a dedicated phone number you can call to interview its Intelligent Virtual Agent (IVA) for a job. With this technology, it is possible to imagine a contact center without a queue. According to the CEO of that company, the goal of this technology is to take the pain out of the voice channel. That pain comes from long hold times punctuated by the message that our call is important, and when we finally get connected to an agent, we have to answer the same questions we’ve already told the system using our voice or keypad. If a contact center can provide a system that answers immediately, has many of the answers for customers, and can hand off to a human quickly and seamlessly (the thinking goes) and voice communication for the contact center becomes painless.One of the most impressive demonstrations I saw involved real-time AI-powered analytics applied to a conversation with a customer. As the agent worked with the customer, the system was displaying a predicted Net Promoter® Score. The score rose and fell as the conversation continued. Now, I don’t know how accurate the score prediction capability is, but given the sophistication of the systems this company provides, I suspect it is quite good.
On the plus side, systems like this could one day eliminate those pesky surveys we are constantly being asked to take. NPS™ scores would also be more accurate because:
- Every contact would be scored, not just survey-takers
- Sentiment would be captured in real time—not days or weeks later
- The subtle but omnipresent bias introduced by survey takers trying to guess what the response should be would disappear
This brings us to the concern mentioned in the title of this article: What happens when we eliminate the customer from Customer Experience?
Is that NPS score displayed by the system really a record of the customer’s perceptions and feelings, or is it a machine-generated facsimile? Since—recent controversy at Google notwithstanding—machines don’t have emotions, I have to say yes, it’s a facsimile. What’s more important is that it is a facsimile created by humans who design the algorithms and feed training to the AI. The complexity is hard to imagine. The words I use when I am getting upset might not be the same words you use. My tone of voice may vary widely from yours or anyone else’s when I’m working up to being upset or angry. (I personally know people who get very quiet when they are upset, and some who raise their voices.)
It seems to me that we are asking a lot of a technology that has a tough time dealing with context. Last October I wrote this Journal entry that pointed to one of the problems: My GPS changed my route based on an accident 170 miles ahead! It was long cleared from the highway by the time I arrived, but the backend intelligent systems didn’t see it that way. We tend to forget that AI is still in its infancy.
I firmly believe that Artificial Intelligence holds tremendous promise to improve the ways companies interact with customers. I also think that if we are going to give up the emotional components of Customer Experience, we are no longer talking about Customer Experience as we have been defining it.
Roy Atkinson is one of the most recognized thought leaders in IT, service management, and customer experience. He is a prolific writer, speaker, webinar presenter, and podcaster as well as an industry analyst. His expertise has been featured by The Economist, BizTech Magazine, Social Media Today, Computerworld, Oracle Customer Experience, SAP Business Innovation, and others. He was described on CIO Insight as a “model for the future digital leader” and by Nextiva as one of the “Top 50 Customer Service Experts of the Decade 2010-2020.” He was HDI’s 2019 Lifetime Achievement Award honoree. He holds a master’s certificate in advanced management strategy from Tulane University’s Freeman School of Business.
As always, you hit the nails right on their heads! The challenge here continues to be the human understanding of the support experience and designing it to be better rather than shifting the same bad experience to a new channel. That takes understanding of that experience and having a support strategy. This is missing in your example of the poor experience of keying in information then having to repeat it to the human. Bad design will perpetuate this in the AI world.
Thank you for your thoughtful comment, Phyllis. This why it’s so important to optimize before you automate and to pay attention to how real users act in the real world, which may be different from our expectations.