Why all chatbots are not created equal

Home/Banking Strategies/Why all chatbots are not created equal

Why all chatbots are not created equal

The promise of automated self-service is an alluring one for financial institutions, though the reality of true support automation too often becomes bogged down by inefficient and cost prohibitive barriers.

As the industry accelerates towards a digital-first environment, customer expectations fast-track alongside it. These customers have been trained by Amazon, Uber and Netflix on the type of experience they should have with businesses, and their banking relationship is not exempt. Customers want to self-serve – they don’t want to call a contact center to ask about routing numbers and many don’t want to go into a branch to open a new account.

When engineered and deployed thoughtfully, chatbots can be an intersection where business value and customer expectations meet. Banks have reduced support costs and the potential to grow deposits and loans, all while meeting customers at the time and channel on which they want to communicate.

Simply automating the answers to simple inquiries such as routing numbers, hours of operation and password resets can free up contact centers for higher-value work. Guided conversations can address more complex situations such as account openings, credit card concerns and loan information. Connectors and APIs can facilitate live chat, appointment setting and other tasks.

However, for almost all banks and credit unions, implementing a chatbot that truly delivers support automation has not been a fast or cost-effective solution.

The reality of implementing a successful chatbot extends well beyond the technological mechanism – technology alone will not solve support and knowledge management challenges. A successful banking chatbot requires content that not only answers questions, but engages consumers in a journey and leverages a proven process that ensures the experience is constantly improving.

Banking-specific content

Effective chatbots require a lot of content. It is the fuel that drives not only the customer experience, but the ability for the chatbot to deliver relevant answers and become more intelligent each time. This is the step at which most financial institutions get stuck.

Building out multi-step conversations to facilitate common customer requests requires complex decision trees that need a banking-focused skill set to ensure customers can easily and logically move from one step to the next. It also requires a deep understanding of not only how to answer the questions, but why customers ask the question in the first place. In other words, how you anticipate what they want to do next and how you tee that up for them within the chatbot conversation.

Content for a banking chatbot is often not a one-time investment, but rather it’s a starting point that requires continuous attention and fine-tuning. It requires analytics to deliver insights into what information your customers are looking for, where customers drop off, and what additional content is required or needs optimization. A seemingly simple chatbot conversation, such as “open an account” or “reset password,” can have well over 25 decision trees.

A process to prevent ‘bot rot’

Chatbots are not set-it-and-forget-it tools. They require consistent analysis, tweaking and optimization to ensure they are constantly getting smarter and delivering a better experience.

Artificial intelligence generally refers to the ability to take inputs, such as sentiment analysis and machine learning, to gauge the top questions being asked and where consumers might be abandoning the process, and then use those inputs to improve the chatbot. But the part that is often overlooked is that it also requires a lot of human intervention to test the content and decision trees to see what works and how the chatbot can be improved.

Creating a banking chatbot that delivers on the promise of automation has its own set of challenges within this highly regulated industry. While sentiment analysis and natural language processing have become table stakes for today’s chatbots, banks must also look for a solution that ensures compliance. To this end, an effective chatbot must include knowledge management functionality such as:

  • Full version control to see when a content item was changed, approved and by whom
  • Content approval workflows that ensure any new or updated content is going through the proper approval process,
  • Bank-grade security with full SOC 2 Type 2 audit for recordkeeping
  • The industry connectors that allow your consumers to continue their journeys from application to application.

Chatbots, like all new technology, are generally misunderstood. The industry is saturated with content from fintechs and early adopters showcasing use cases of how consumers can use chatbots to check their balance and pay bills.

The reality, however, is that banking consumers don’t need another tool for transactions. They have mobile banking and online banking that allow them to easily do that today. What banking consumers need is a tool in your digital channels – web, mobile, online banking – that provides support and service. Chatbots need to provide basic support with products answers and intuitive journeys that guide your consumers to the next step.

DJ Haskins is vice president of marketing at SilverCloud.

Want more BAI Banking Strategies? Sign up for our free newsletter.