Every December for the last five years, I’ve compiled a set of predictions about what is likely to happen in the customer experience space in the coming year.
Last year, one of the themes that emerged was that the contact center and customer service, in general, would lead the way in the application of generative AI. The prediction mooted that this technology was set to transform not only customer service agent productivity and efficiency but also the self-service tools that brands could deploy to allow customers to resolve a large number of their own queries and problems.
Many technology vendors were very bullish about this latter opportunity, with some promising that their technology would help resolve 80% of customers’ queries straight ‘out of the box.’
The reality, however, is very different.
Recent research from Gartner finds that while 73% of customers indicate that they have used and want to use self-service tools, only 14% were able to solve their issue using the tools that were available, however, for “very simple” problems that number improves to 36%.
Speaking to Paul Adams, Intercom’s Chief Product Officer, in the run-up to their annual customer event, Pioneer, which took place on Thursday last week, he believes that many of the claims being made by vendors are either ‘bogus’ marketing or that they are ‘cherry-picking’ statistics from their clients’ experiences and then generalizing them in their marketing.
Intercom is trying to buck this trend of inflated promises by being open with what they are doing and the results that all of their clients are achieving as a way of injecting some transparency into the space. With that in mind, they announced at Pioneer the latest version of their customer service-focused AI engine, Fin 2, and shared how it can, on average, achieve a resolution rate of 51% ‘out-of-the-box’ with a 99.9% accuracy rate. This is up from the 23% average resolution rate that Fin 1 was able to achieve ‘out-of-the-box.’
In his presentation at Pioneer, Adams shared a distribution chart showing the resolution rates that all of their Fin 2 clients were achieving and emphasised that the 51% number was only an average and that many of their customers were achieving resolution rates at much higher percentages.
Adams explained that achieving such results has been a long and arduous process of improvement involving hundreds of A/B tests over the last couple of years. One of the most significant ways that they have been able to achieve this is not only through the use of the best Large Language Models (LLMs) on the market in combination with their AI engine that has been custom-built for customer service but also because they have worked really hard at the front end to build a product that better understands the questions being asked of it.
This is significant as understanding a customer’s initial question is a common point of failure with many automation and self-service tools. Research from Gartner has found that 45% of customers failed to resolve their issues using self-service tools because the company didn’t understand what they were trying to do. Eric Keller, Senior Director, Research in the Gartner Customer Service & Support Practice, adds that “Self-service can offer substantial benefits for organizations and customers, but work is required to ensure that customers’ needs are understood and responded to.”
However, while Intercom’s clients’ results are impressive and their approach to transparency is laudable, what I found really interesting when talking to Adams were two of the client stories that he told me.
The first story was about a customer who has always struggled to hire the right people for their customer support team. However, they’ve now been able to automate a significant amount of the human support they were previously giving to their customers, which has allowed them to stop looking for new people. Moreover, it has freed up so much time and space for their customer support team that they are now able to do things like turn on their phone channel, which historically was always deemed too expensive, and focus their team on activities like customer success or more expansionary conversations with customers that involved an element of selling.
The second story was about a customer who initially was dead set against turning on artificial intelligence within their customer service environment as they prided themselves on their human support, believing that artificial intelligence could never be as good as their “wonderful humans.” However, over the course of one weekend, they faced an emergency and a huge surge in customer queries. In desperation, they turned on Intercom’s technology in the hope that it would provide some help for their customers but also some respite for their agents. To their surprise, and after QA’ing the responses that customers received, they found that it was as “good as our team at Tier-One support.”
These stories are fascinating examples of, on the one hand, some of the real concerns that many customers have with implementing artificial intelligence into their customer service environments and, on the other hand, what it allows others to do when they start to realise the benefits.
Overall, the big lessons for me that came out of my discussion with Adams and Intercom’s Pioneer event were:
The market is full of big promises, so brands’ biggest challenge is separating the signal from the noise.
This technology is getting better and better all of the time, so even the most sceptical brands should be investigating its possibilities.
Finally, while some brands will use this technology to reduce their reliance on human agents and cut costs, the more progressive brands and those that believe that great customer service teams can be a real source of differentiation will leverage this technology to allow them to deploy their team’s experience and expertise in places and situations where they can really add value to both their customers and the business as a whole.