When chatbots fail
by Mariette Johnson Wharton
Photo Credit: EKR Agency
The Power of Personalization in Banking Report shows some interesting and (what should be) fear-inducing trends in financial institutions. First, financial institutions overestimate the value that customers place on their banking relationships. Fatal mistake if this leads banks to offer too much unsolicited advice. It would be creepy for customers to recognize how much data banks have about them, for starters. Secondly, over-meddling is annoying.
What should be cautionary is that the report’s findings indicate that the majority of all kinds of financial institutions are “unprepared to provide personalized advice, communication and offers.” So, while the goal of personalization is certainly laudable, you need to know what you are doing or your strategy will backfire.
Personalization through technology is the latest for financial institutions and others looking for high-touch customer experiences. All fine and good but care needs to be taken not to annoy customers with messages selling the benefits of the latest bank offerings. Customer-initiated messaging is one thing and even notifications about potential problems would be welcomed but sales pitches, probably no. TakePersonetics’ marketing language about its AI abilities:
"Using real-time predictive analytics, Personetics Anywhere enables financial institutions to go beyond simply reacting to customer requests. By anticipating customer needs and delivering personalized and timely insights, banks can help customers be better informed and more confident in their ability to control their financial affairs. For example, a customer that is traveling to a different country may receive advice on how to use the bank’s card in that country while minimizing any associated fees. Personetics Anywhere can also analyze customer spending patterns and provide advance warning if an account may become overdrawn, suggest ways to increase savings, and deliver other insights related to the customer’s most current financial activities."
Information about fees could be ok, and doubtless, warnings about an overdrawn account would be welcomed, but beware overstepping boundaries and allowing a bot to irritate customers. Suggestions on increasing savings veers into dangerous territory and frankly, it’s unclear what “deliver other insights related to the customer’s most current financial activities” might be, but alarm bells are starting to go off.
Worse than being unprepared, AI chatbots cannot be counted on to deliver appropriate responses without some serious filtering. Machine learning technology is at an early enough stage that even giant Microsoft, with its nearly limitless resources, has found its attempt at a chatbot, teenage girl emulator Tay, has fallen prey to the teachings of Internet trolls, resulting in @TayandYou tweeting shockingly racist statements in support of genocide. As an experimental project, it will get chalked up to learning but tweets of that nature in a customer relations scenario would cause irreparable brand damage to say the least.
Facebook Messenger’s AI chatbot, targeted at businesses, is expected to improve customer engagement by being able to send links and images and answer questions and will hopefully not fall victim to inappropriate machine learning. Google’s AlphaGo has already shown the promise of deep neural learning, as has IBM’s Watson. Bots might be able to win strategy games against the world’s most competitive players, but they still lack the ingenuity of humans so even if you think we can overcome the Tay fail with filters, it can’t bring the creativity of a person. At least not yet.












