Given how enterprises feel about the importance of customer experience management (CEM), and given the current obsession with AI, it’s no surprise that most enterprises say there’s a role for AI in CEM. The problem is that while over 90% of enterprises tell me they are “evaluating” AI in CEM, less than ten percent have actually taken this evaluation process to the level of testing AI. This, despite the fact that major vendors like Cisco and Juniper have already incorporated AI into network management. Why is there this disconnect, and what does that disconnect mean for both CEM and AI?
I said in my last blog that 248 out of 341 enterprises who had offered me views on customer experience (CX) said their CEM plans and tools didn’t fully recognize the impact of the Internet and online information on their customers. This was the largest issue enterprises cited with CX/CEM, in fact, so if there’s an AI mission in CEM, being able to manage the modern, online customer experience would be a worthy target.
If networks are delivering the critical elements of the customer experience, why isn’t AI in network management automatically engaging AI in CEM?
That, at least, is a pretty easy question to answer. The network is a conduit for delivery, not the content. Every enterprise application that relates in any way to CX is just that – an application. You can have a perfect network connection to an imperfect application, or your application can be experiencing its own hosting problems, and it does not matter. The user does not care – they will still have a bad experience. In fact, of the 248 users who said online impact on CEM wasn’t being fully addressed, 161 said that application problems were the biggest issue, where only 76 said it was the network. Of the 44 who said they used AI in network management, only 18 cited that use as a source of CEM improvement, obviously because the network isn’t the whole picture. Enterprises are starting to recognize this, and 76 of the 248 had at least an idea they were pursuing. The question is whether any of the ideas will help.
The most popular notion on the use of AI in CEM is to use it to address user problems, meaning an AI chatbot. While 33 of the 76 enterprises with an idea of applying AI to CEM cited this, 187 of the group who saw a CEM shortfall in online CX said it was a bad idea. They said that offering a solution to a problem a customer was experiencing wasn’t CEM at all, in fact, and I think they’re right. In fact, enterprises overall think you have to prevent problems, not respond to them.
That’s also the view of 14 of the 76 enterprises with an AI strategy for CEM in mind. Their thought was to use AI but to analyze customers’ chat records to detect patterns both in the sources of customer complaints and the responses that most satisfied customers. This information would then be used in two ways — to reduce the issues through changes in applications or even products and to train online support personnel and retune support applications.
The broader enterprise community seem to be at least interested in this AI approach, but uncertain about how it might be implemented. First, they were unsure whether AI could be trained to analyze the support chats, but they were most concerned about how their organization could act on “changes in applications or even products” because this clearly would fall outside something AI could do. In the end, it was felt using AI to identify “changes in applications or even products” had little value beyond what a review with customer service personnel could deliver.
Idea number two for the use of AI in CEM focuses on the entire online experience. Instead of having AI manage only the network, you have AI manage, or at least analyze, the online quality of experience (QoE) for all the customer-facing applications. If QoE goals are set properly, and if AI could manage networks and hosting to achieve these goals, then CEM would surely improve. In all, 28 of the 76 users with an idea of how to apply AI cited this idea as their target, and 54 thought it might be useful.
The problem with this is the sheer magnitude of change required. None of the original group of 341 enterprises said they were using AI to manage application hosting, and only 116 suggested they even knew of a way to do that. Further, 221 of the 341 enterprises indicated that CEM for today’s online-centric world had to do more than resolve issues with overall application QoE. In short, fault and performance management even for the whole network/hosting ecosystem didn’t guarantee a good CX.
The third approach that was being considered was to use AI to analyze the paths customers took through the website and the cloud/data-center applications related to it. Here the goal was not to alter applications, but to use trajectory data, and in particular “abandonment” data to determine wher4e information presentation could be improved. The theory is that you can infer, by how customers navigate product/service information and support help/knowledge bases, whether they are actually seeing what they want and need to see.
This idea, which had support of only 15 of the 76, actually seems the one most likely to be valuable if you analyze what enterprises think about optimizing CEM with AI. Unless the “experience” an enterprise is trying to optimize is online content delivery or shopping, there’s too much of the customer experience that falls outside network/hosting management for the first or second ideas to help much. Analyzing customer access to online data, on the other hand, gives you a picture of how your information is being used, and that gives you a picture of what a customer is trying to do, and whether they’re successful.
A shift to customer relationships built through online interactions depersonalizes the customer-vendor relationship, but this same shift gives a company the potential to use accumulated data to look inside the mind of a customer or prospect as they consider, or commit, to a purchase. It’s also a mission that AI could clearly support, though it would require tracking how an online user navigated through material. It’s disappointing that this AI strategy had the least current support, because it’s the most promising.
