BoA chatbot uses continues to soar…
“Erica,” a Bank of America chatbot rolled out in June 2018, has hit the 10 million user milestone and is on track to complete a huge 100 million client requests in the coming weeks, the BoA — which serves 66 million retail/small business clients — said today.
Development of the Bank of America chatbot started in Autumn 2016, when the BoA pooled together a 100-strong internal team dedicated to building up the voice- and chat-driven tool, which uses AI-powered predictive analytics and cognitive messaging.
As of May 2019 there were over 400,000 unique ways that clients could ask Erica financial questions, with the number continuing to grow, the bank said [pdf] and appears to have secured impressive user-growth and back-end integration.
Its success has focussed industry minds on comparable tools globally: amid macroeconomic and regulatory headwinds, banks are under strong pressure to reduce high customer service overheads and chatbots are seen as a key way to do so.
Today’s milestones came as the bank rolled out a range of new Erica functions, including a proactive prompt when a a merchant refund is made.
David Tyrie, head of advanced solutions and digital banking at Bank of America said the combination of the technology and a “high-touch approach” delivers a “more intuitive and efficient banking experience for our clients across all channels.”
During the third quarter of 2019, digital clients logged into their BoA accounts 2 billion times and used digital to make 138 million bill payments, the bank said.
R&D work to improve conversational AI capabilities for chatbot-like tools continues apace. Earlier this year, for example, Microsoft open sourced a toolkit called Icecaps to help developers “imbue their chatbots with different personas”.
That toolkit is a TensorFlow-based, modular framework designed to make it easier for users to create sophisticated conversational AI training configurations, using neural networks that involve new signal processing methods and deep learning algorithms.