Page 18 - EngineerIT September 2022
P. 18

ICT


                    Why can’t your chatbot just


                                     solve my query?






                                            By Ryan Falkenberg, Co-CEO, CLEVVA




             hatbots are, to be blunt, everywhere. It feels as if you can hardly visit a website
             these days without a bot asking how they can help you. It’s not just a matter of
       Cperception either. According to research from Gartner, chatbots will become a
        primary customer service channel for around a quarter of organisations by 2027.
           Ubiquitous as chatbots are becoming, however, they still have a significant problem:
        aside from their inability to understand customer queries, many of them simply aren’t
        that helpful. Many of us know all too well how frustrating it can be to type multiple
        variations of a query into a chat box, only for the chatbot to keep returning irrelevant
        results (especially if there are no other avenues to contact the organisation in question).
           Frustrating as it is that the typical chatbot is so poor at solving queries, it’s
        understandable. Fortunately, there is an approach that can make them much more
        useful.
        The trouble with chatbots

        When it first became clear that chatbots might be useful to the world of business   Ryan Falkenberg
        (sometime in the mid-2010s), many organisations vastly overestimated what they’d
        be capable of. In their minds, they were getting a like-for-like replacement for human
        customer service agents. Chatbots, the thinking went, would be able to resolve   Additionally, the resulting actions can
        customer queries completely unassisted.                                   be processed with confidence because
           In reality, what they got was more like a smart librarian. That is, the       they’re more likely to be right.
        chatbot doesn’t know the answer itself, but knows where to fetch it from.   A Gartner   That in turn means that
        Unfortunately, many of them are working with poorly constructed                        organisations are more
        knowledge bases, meaning that the answers they provide tend to be   customer service and   willing to embrace
        generic and unhelpful (much like a librarian who only has access to   support (CSS) survey of 50   automation and the
        out-of-date reference books).                            respondents conducted online     benefits it brings.
           It’s hardly surprising then that interactions with these chatbots   in January and February 2022
        can be so frustrating, especially once they go beyond simple   revealed 54% of respondents are   Steps in the
        requests such as “buy airtime” or “milk tart recipe”.   using some form of chatbot, VCA    right direction
                                                                   or other conversational AI     While many of
        The service bot solution                                                                  today’s chatbots
        Fortunately, there is a relatively simple solution when it comes to   platform for customer-facing   aren’t the customer
        overcoming the limitations of chatbots. It involves adding a single   applications     service panacea
        additional step to clarify the intent of your query. Adding this step and           organisations imagined
        doing so usefully requires the integration of service bots. A service bot is    they would be a few years ago,
        essentially a digital service expert that can help clarify your request, analyse your   improvement is possible. With the help
        needs, identify the root cause of your problem, and recommend a solution.  of service bots, most organisations can
           That’s important for two reasons. The first is that it ensures that you start in the   put themselves on the path to offering
        right place. So, for example, if you type “SSL Certificate” into a service bot, it can   customers a Virtual Agent - one capable
        clarify whether you want to update it or if you want to get one. The more of these   of resolving their queries first time.   n
        clarifications it goes through, the better its natural language understanding (NLU) gets,
        further improving its ability to answer queries.                          CLEVVA specialises in building, deploying
           The second is that you’re adding the capability to analyse a problem and come   and maintaining Virtual Agents for
        up with a relevant solution. Most chatbots don’t do that. They jump to a solution   regulated industries. These customer
        set immediately, which is not helpful if you don’t know what your problem is.   service experts drive down the cost to
        With analysis, the resolutions the customer’s presented with are much more   serve while improving service compliance
        appropriate.                                                              and customer loyalty. www.clevva.com


                                                 EngineerIT | September 2022 | 16
   13   14   15   16   17   18   19   20   21   22   23