Starting a banking virtual assistant prototype seemed like a technical challenge, finding the right platform, understand connection with common messenger channels, finding the right underlying programming language. Multiple options are available and after some investigation I decided to go for the Amazon Lex platform. Reason for that is that you can program so called Lambda functions in Python which is easier for me compared to other languages.

Once done the few tests, the challenge is to find the right dialogue flow, this is the interesting part. How to make a dialogue flexible, natural and not just a simple replica of standard forms.

First start designing the flow according to the user intent, for example let the user determine what he wants to do, opening an account, buy a car or start trading on the stick with exchange. You as a dialogue designer needs to understand what a user is likely wanting to do. We could just ask with a few options, but that is not the natural way people think. Most probably some do know what kind of product they want from a bank, but some don’t. You have to cover both cases and make the dialogue flexible. This is the hard part, one user may say I want to open an account, other may say they want to buy a car and some will say I want to invest 20k in Apple shares.

Although the technology is still evolving, the understanding of human behavior and way of thinkong is even more critical. So far I have not seen banking bots offering such flexibility.


(Visited 20 times, 1 visits today)
%d bloggers like this: