How many times have you asked a chatbot to cut the clutter and direct your query to a customer service agent? Or, have you avoided the bot as soon as it lost its vanity? How many of us continue using Alexa Echo or Google Home extensively? Most of us would reply in the negative.
The industry has often complained that the chatbots have a success rate of less than 20%. It’s disheartening to see what could have been a popular mode of banking engagement fail so miserably. The technology is marvellous; the engineers have done the job right. It can work flawlessly, yet despite all this it disappoints most of the time. Where did the industry go wrong?
Alan Turing once said, “A computer would deserve to be called intelligent if it could deceive a human into believing that it was human.” Our customers returned home disappointed because they expected nothing less from our bots. The technology, the digital assistant programs, and translation services all work flawlessly, yet our bots disappoint.
This is because we failed our bots! We didn’t teach them sufficient skills to go online and face the world. A bot is only as good as the skills you teach them, and any intent detected outside the defined skill sets leaves the bot helpless, and the customer frustrated.
Most of the banks and software companies took the approach of attacking the most popular and used transactions first. Asking chatbot for account balance, recent debits, pending utility bills sound cool. It brings out the real capabilities of the system. But then, were these the skills that the market was ever looking for on their Digital Personal Assistants (DPAs)? In the course of the digital revolution, banks have brought in many new touchpoints for their customers. Most of them have the same or similar capabilities. From SMS banking to in-app banking keyboards, all touchpoints flawlessly execute inquires and transactions. All these options are available to the customer as single touch banking options. Did the customer need another channel to perform these tasks?
The user of today learned and adapted to multiple channel options and navigated around in the absence of one from the other. They don’t land at the branch or dial customer services for these menial tasks. They seek to connect with the bank only when they have a query or need to report an issue that’s not possible using self-service channels. This is where they need a human. For a channel that enables two-way interaction and emulates humans, it’s evident that the customer would want it to do more. Bots required better skills.
This is where banks need to have a deep understanding of their customers. A one-size-fits-all approach does not work anymore. Banks need to define intents by analyzing the insights collected from user behaviors. These intents should then be converted into skills that our bots can acquire. The tasks that a customer expects a bot to do is very different from the show-and-tell tasks we taught our bots.
Banks have years of expertise and customer relationships to understand the needs of their customers. The user segments, customer expectations, and brand values are best understood and preserved by banks. Interactions with customers give them an edge in deciding the features most needed by their customers. They are best placed to understand customer intents. Banks need to ensure that they have a process to capture these intents, validate them for a customer segment, and build them into their chatbots.
The success of a chatbot is dependent on three interconnected cog wheels that need to work in unison; Artificial Intelligence, Training, and APIs. Artificial Intelligence is derived from the Digital Personal Assistant (DPA) Platform that the bank implements. The new generation DPA platforms allow the bank to have a footprint across multiple interaction platforms ranging like Facebook Messenger, Whatsapp, Wechat, Amazon Echo, Google Home, in addition to bank’s customer portals. These engines interpret a customer’s communication and derive the intent of the conversation. Once the customer’s intent is derived, the skill of the DPA takes over. It invokes the functions to process the intent and logically map them into name-entities for downstream processing. The ability of DPA to derive the intent of a customer require much training and is a continuous process. Banks would need to ensure that their administrators investigate the conversations. Administrators also need to observe exceptions/fallouts and train the bots to sharpen the skills configured to make them consistent, cognitive, and congruent. All bot programs and DPA engines allow for quick configurations of skills and training, enabling the first two cog wheels.
The last and most crucial cog wheel is APIs. These are the lifeline of a chatbot and enable its skills. These APIs make or break a chatbot experience and hence need to be reliable, robust, and granular. Banks need to publish APIs at scale, manage entitlements, and authorizations of these APIs. They also need an intelligent engine that can connect and navigate across the product processors without causing a shift in the bank’s network and application architecture.
Subscribe to our Blogs:
Oracle Financial Services Blogs: Sign up today