On average, 60% of American adults spend their time online daily. Imagining that during the conversation with your friend, you need to perform some mobile banking activities. It will be efficient that you could just conduct all activities within Facebook Messenger.
What it does
With Rory, you could check your current balance, pay your friends, check the expense of last month and obtain customized financial advice.
How I built it
We integrate with Google Dialogueflow.io into facebook API to build the AI assistant. The backend is constructed with Flask framework in Python. All finastra apis are called within the Flask framework in sequence. The personalized suggestions and analysis are generated by numpy, matplotlib and sklearn libraries in Python.
Challenges I ran into
First of all, It was hard for us to define the problem and build a nice use case around it. Then, the framework and APIs are new for all of us. It takes us a significant amount of effort to integrate different APIs and deploy the app in Heroku. Some tricks such as returning graphs, returning text with graphs and extract parameters in Dialogueflow.io require us to play around and creatively achieve it. Eventually, we manage to deploy and run the App in a stable environment.
Accomplishments that I'm proud of
- A well-functioning app that works in Facebook Messenger
- A nice use case around the app
- A disruptive idea for retail banking that captures Millenials lifestyles
What I learned
- Gained knowledge about pain point identification and design of user journey
- Smooth communication and cooperation techniques within the team
- Technical capabilities
What's next for Financial Assistant on FB Messenger
- Integrate with Oracle SQL database that could store the context of the conversation for future reference
- Designed and aggregate more features for the AI assistant 3.Build a minimum viable product based on the current POC