Inspiration
When we go to bank, for most of the non trivial work, I see the employees asking their seniors or call people in other branches/head office for the process. Though most of these are documented, finding them is pretty difficult. Also, searching for relevant documentation is another pain
What it does
Banks, being one of the consumer heavy industry, with most of the customer facing employees not very well verse with handling software technology, a messenger kind of chat with nlp integrated, eases lives of many bankers.
What we have presented is a chatbot for the employees, which employees can interact using natural language. We have presented 2 kinds of workflows.
- Account creation workflow, where a BPMN (Business Process Modelling Notation) diagram is showed to the employee. As bank or other businesses are more familiar with BPMN diagrams, this is displayed to the user
- Interest rate inquiry workflow, where there is not much people interactions are involved. This is displayed as QA session, where employee keeps on entering the details for which information is required and information displayed gets narrower and narrower
The workflows, that are implemented in this prototype are:
- Account creation
- Amount transfer
- Loan grant process
- Interest rate inquiry
First 3 workflows displays corresponding BPMN diagrams and 4th one is a series of message interactions
How we built it
We used Facebook's messaging infrastructure to implement this. For NLP based queries, we used defaults queries. We also added few entities in wit.ai to support custom NLP queries. The web hook in hosted in Heroku.
Accomplishments that we're proud of
For the message interactions, we created a workflow diagram, using which we created a config file implementing the workflow. Hence, in order to add new workflows, one need not know how the code is implemented. Just create a tree in our format, if new wit.ai training is required, create and train an entity there. In this way, minimal code changes are required for any workflow addition. We could also externalize this workflow creation, so that the core code is not touched
What we learned
- Basics of wit.ai
- Messenger creation and interaction with a Facebook page
We got to know that once we understood how this works, writing an application is pretty easy with minimal code
What's next for BankTutor
We are planning to create a full blown application, where addition of different workflows is made as easy as possible, so that any admin in a bank can use it with ease
Links:
- Code in Github: https://github.com/ksholla20/bankTutor-messenger-webhook
- Webhook on Heroku: https://pure-cliffs-78206.herokuapp.com/
- Messenger Link: https://www.facebook.com/messages/t/575336899760839

Log in or sign up for Devpost to join the conversation.