Customer service and conversational AI – when to use it and when to avoid it for the best customer experience


Share on LinkedIn

When considering whether to automate a significant part of customer service with a robot, there’s only one thing that really matters: what is plan B?

The complete customer experience should always be put first. No matter how robust the natural language understanding your new virtual agent is built on, customers have a tendency to trip up even the best machines. Issues could arise for a number of reasons; from questions out of scope, unique unforeseen ways to structure a question, to outright refusal from customers to use your brand-new robot. It will happen, and it will happen regularly. The crucial part is considering what happens next.

If you care about customer experience, nothing is more important when automating customer interactions than the virtual agent’s ability to know when it doesn’t know. What happens when it gets a question or request it doesn’t know with certainty how to handle? It’s usually one of two things: it either admits that it doesn’t understand, or it provides the wrong answer. But what if there was a third outcome?

Take for example DNB, Norway’s largest bank. They successfully resolved a high percentage (51%) of their incoming customer enquiries. But it isn’t smart to focus just on the 51% that get their enquiries resolved. You should also care about the remaining 49%. What happens with their requests? In DNB’s case, the remaining 49% are seamlessly forwarded to the right person in the customer service centre. The human agents quickly read the chat log and provide the customer with the help it needs in a matter of seconds. It’s all thanks to how they positioned their virtual agent, Aino, who acts as a first responder and a gateway to human customer support.

The move has provided them with a mind-blowing 50% drop in chat traffic to human support, and has resulted in reduced wait times, on top of their now instant customer service, to significantly enhance the customer experience. With approximately 20,000 online customer interactions every day, the scope of their customer interaction automation is probably one of the best in the world – in any sector. Aino was launched in October last year and they’re already working to add more transactional abilities; so DNB is just getting started.

So, automating for automation’s sake is not a good plan. Make sure it’s a move that improves the customer experience. Humans and machines really do make an incredible team under the right circumstances.

More than just “AI”

AI does not (yet) represent the sentient, self-learning machines from science fiction. Just like its human colleagues, the successful AI-powered digital employee has some specific requirements. Those requirements are not substantively different to those expected from a human, but they have a very definite AI twist.

● Extensive industry-specific knowledge and experience – Our clients possess a huge amount of industry knowledge and experience which is very specific to their business. If their new digital employee is to be successful in their new job, they will need to rely heavily on this intelligence. Users will often see through attempts by third parties to recreate this information on behalf of a client, and it will almost certainly detract from their experience of interacting with the virtual agent. Taking time and care to transfer high-quality industry data, ensures total uniformity in the answers the virtual agent provides to end users – a feat not easily replicated by their human counterparts.

● Organizational intelligence – For a virtual agent to authentically represent a client’s brand, it requires an understanding of their core values and goals. Additionally, but of no less importance, is the need for it to understand and represent the corporate memory, manifested in a deep understanding of the history of the company and its unique selling points. By making certain this knowledge is accurately transmitted, clients can be assured of continued best-in-class customer service from their new virtual agent.

● Outstanding customer care – When developed and trained correctly the virtual agent’s customer care abilities should at least mirror the service its human colleagues provide. When digital and human service becomes more indiscernible, customer satisfaction increases.

● Continuous improvement – A static virtual agent runs the risk of very quickly causing the end user to lose interest. Replies need to be kept current and intents should be frequently refreshed to ensure that information about the client, and the products or services they provide, continues to be relevant. It is also important to constantly analyze data from user interactions to make sure that the virtual agent is meeting their ever-evolving needs.

Conversational AI is now at a place where it is mature enough to achieve the right balance between speed and accuracy, based on relevant information, conversation history, and predictive analytics. Many businesses are using it to provide the most accurate answers to first-response questions, even as policies, people, and products shift.

Virtual agents also tap data-intelligence to route inquiries to an appropriate human agent when needed and make second-line recommendations about how to handle issues or identify problems when the hand-off occurs.

While some businesses have yet to be convinced of the benefits of AI for handling customers, leaders like DNB and Santander are charging ahead and showing the way. They are driven by not only a desire to increase revenue and reduce costs, but to accelerate the customer experience in the very first instance. And they have proven that significant gains are there for the taking.


Please enter your comment!
Please enter your name here