Zendesk released its annual Customer Experience (CX) Trends Report this week and is using its findings to promote a concept it calls Immersive CX. Does this mean the company wants us all to don VR headsets and plunge into the Metaverse to call up customer service? Fortunately, no — or at least, not yet. I checked in this morning with Matthias Goehler, EMEA CTO at Zendesk, to find out what Immersive CX is all about. Here’s his definition:
What we mean by this is that you as a customer, irrespective of B2B or B2C, you should be able to get in touch and interact with the company in the most natural way you want to do this, [on] whatever channels you prefer.
By the way, yes, this could be the Metaverse in the future, this could be in a 3D digital world, if companies go that way. But it could also be a call, it could be an email, it could be social media, it could be messaging, it could be a chat online. And ideally … you are being seen as an individual, as a customer, not as a ticket or not as a transaction. The company is trying to help you in a personalized way.
Now this is a concept I can get on board with, so long as I can leave the VR headset out of it. It’s in line with my own assessment of today’s digital user experience and its ability to both convey information that’s relevant to the user’s context and at the same time gather and analyze data about their experience. It also appeals to customers, according to the CX Trends report, which finds that “61% of customers are excited about experiences that are natural, convenient and fluid.” To be honest, that’s stating the obvious, but it’s in stark contrast to the awkward, cumbersome and stilted experience that characterizes all too many customer interactions. The report also finds that businesses realize they have to up their game in this regard, with 71% of leaders “committed to reimagining customer service.” The full CX Trends Report is available for download from Zendesk.
Inevitably, the message Zendesk wants readers of the report to take away is the need to invest in better CX processes, preferably using all of the cross-channel personalization and AI-fueled automation that its platform promises. But the history of CX is littered with case studies where increasing digital automation has ended up producing worse outcomes for customers rather than making things better. Much of this has been the result of trying to eliminate human interaction altogether, rather than finding the right balance between machines and people. How to avoid the same errors when rolling out AI and machine learning was the main topic of my conversation with Goehler. He walked me through three separate areas where the technology is being applied.
Automating customer interactions
The first of these is in support of interactions with customers. This is where bringing the right contextual information together is so crucial to a personalized experience. He says:
I might have started an interaction in the digital world, be it online, or be it in 3D, whatever. Then if I go to the store, it would be great if the store associate had knowledge about my previous interactions. Then I don’t need to repeat myself and I don’t need to answer the same questions. They have the history, they can help me in this moment. How you manage all of these different channels and how you make it seamless and fluid between those, will define a great customer experience.
When it comes to automating online interactions that have previously been handled by human agents, it’s important to work with agents to make sure the process runs smoothly. He shared the story of an IT and telecoms provider in the Nordics that wanted to improve its handling of commonplace customer service requests, such as changing a rate plan or canceling an account. When it asked agents how they carried out these tasks, it discovered there were several different processes in use. The first step was to work with the agents to standardize on a single process before going on to automate it. The process also had to accommodate integrations to other systems such as billing and customer account records, and to handle exceptions, such as a customer having more than one account. All this preparation meant that, when the system was rolled out, it was welcomed by agents as well as customers. He says:
It’s obviously super convenient for me [as a customer]. I can do this whenever I want, outside business hours, whenever I think about it. It goes relatively fast — I don’t need to wait in a queue whatsoever. And I just get it done.
On the agent side, interestingly enough, it improved the agent satisfaction as well. They didn’t fire anybody, they still have the same people. And the agents felt, ‘Great, I can get rid of all of these simple, repetitive tasks, and really concentrate on the more complicated cases’ — perhaps the cases where customers are unhappy, and it’s advisable to talk to a person.
AI supporting agents
AI can also support agents in their work, first of all by evaluating the context of an incoming call, such as intent, language, relevant skills and even sentiment, to route it to the person best placed to handle it. He elaborates:
Internally in your operation, you might consider treating incoming requests from unhappy customers or angry customers differently than the neutral ones, or the happy ones. You might have a different SLA, perhaps you even have teams where the people are better trained and better suited to deal with critical customer situations than others. It’s an important piece of information that you can get out of the request.
We can do this all automatically before the agent has even seen the incoming request, then as a second step we can route it to the right person.
When the agent answers the call, the AI can select and present relevant information to help resolve it. That might be the order history if it’s a return, the most relevant knowledgebase articles to answer a specific question, or a suitable selection of predefined macro statements or actions. He explains:
In the combination of knowledgebase, macros, relevant information, ideally, we make you much more effective. You don’t need to search for this yourself. You get the info.
You still answer the request. You’re still interacting with the customer, and you can still decide what to do and how to do it. But we save time, because we pre-populate, we’ve searched for you, et cetera.
The final area where AI plays a role is in behind-the-scenes administration, helping maintain and improve the CX system. For example, it can automatically analyze how agents have responded to calls, and perhaps identify gaps in the knowledgebase, or else flag unused articles that might need to be archived.
But despite the current interest in generative AI tools such as chatGPT, Zendesk isn’t suggesting that AI can go as far as compiling the response to the customer — not yet, anyway. The current technology can only create very generic answers, he explains. For example, it can explain why flight delays happen in general, but it’s not able to discover what happened to delay a specific customer’s flight and suggest next steps. That would be an interesting evolution of the technology to watch out for. He says:
The interesting thing to see and to watch out for, is the next evolution of these large language models and the intelligence on top of those, that you now could integrate into the system of a company to move it away from generic to very specific …
Sometimes these answers are interestingly good. But if you can move them out of being generic, and can combine them with operational systems and operational procedures and customer intelligence, et cetera, that will be interesting to see that evolution.
I’m in two minds about Immersive CX as a term. On the one hand, I agree with Zendesk about the potential of a digital UX to help personalize and improve the customer experience, and I think the word immersive does express the extent to which the technology can both instrument the entire experience and deliver rich contextual information very effectively. But I dislike the way the word is so associated with current concepts of the Metaverse, which tend to treat the end user — in this context the customer — as a passive player rather than a proactive participant who has agency to shape their own experience.
One of the big dangers of introducing any form of automation into a customer service environment is that businesses see it as a way of cutting costs rather than an opportunity to improve customer outcomes. We need to keep in mind that, while intelligent automation may be good at reducing the time it takes to resolve the same problem many times over, it typically requires human creativity and judgement to discover the root cause of the problem and stop it happening at all. It’s good to hear Zendesk surfacing examples of where its customers have combined automation with the human skills of customer service agents to engage with their own customers and deliver a better experience.