How to use AI chatbots in banking successfully

This work examines customers’ acceptance of AI chatbots in three different banking contexts: opening a savings account, an investment portfolio, and a pension fund.  The research investigates how the use of AI chatbots is perceived by the customers in each of the scenarios (i.e. how useful, easy to use, trustworthy, and risky the AI chatbot is perceived) and how these perceptions eventually influence the consumers’ attitude towards using the AI chatbot and their behavioral intention to use it.

Relevance: The present research is of high topicality as the use of AI chatbots is growing fast and firms such as banks need to make decisions about how to implement the new technology. For what purposes should a bank use AI chatbots? Which clients should be targeted? Should the banks shift away from human interaction to AI chatbots?  Those decisions are of high financial relevance as they might help to reduce the costs but can on the other hand also impact a firm negatively through loss of trust.

Results: AI chatbots are perceived as more useful, easier to use and trustworthy as well as less risky in the savings account scenario compared to the pension fund scenario. These perceptions strongly mediate the impact on consumers’ attitude towards using the AI chatbots. The attitude towards using the AI chatbot and consumer’ behavioral intention to use it are found to be significantly lower for the pension fund scenario compared to opening a savings account.

Implications for practitioners:

· AI chatbots are well suited for simple, low-risk tasks, where full automation is feasible. This is the case in banking for opening a savings account.

· AI chatbots are not suited  for more complex tasks where risk and trust play an important role. This is the case in banking for opening a pension fund.

· In more complex situations the human interaction keeps playing an important role and AI chatbots should only be used in a complementary way.

· The preference to use an AI chatbot does not depend on the age of the clients, a age-based AI implementation strategy should be avoided.

Method: The method consists of a quantitative online vignette experiment where participants are put into the shoes of customers of a hypothetical bank. The experiment is conducted through the Qualtrics survey platform. A total of 203 participants respond to the questionnaire and evaluate standardized AI chatbot responses in a between-subjects design.  Three scenarios (open a savings account, an investment portfolio and a pension fund) are randomly shown to the participants. The respondents are then asked to evaluate the AI chatbot regarding to different perceptions. The data is analysed using primarily ANOVA and regression analysis by means of the statistics program R.