ARIA
AI representative that will ensure client onboarding experience is smooth and easy.
The idea is to use Generative AI that will simplify document or background checks, predict common actions and avoid redundant steps.
BACKGROUND
This project is part of incubation lab and objective is to explore possibilities of AI to accelerate client services with have quick turn around time, provide personalised information and efficient resolving of client queries.
The Goal
To leverage ability of generative AI to process the previous data and make intelligent decisions.
MY ROLE
I contributed to concept design to represent the user flow starting from document handling to account opening.
DESIGN PROCESS
Discover
Analysis of current user flow and identify opportunities to use Generative AI
Ideate
Elaborate on the proposed user flow and story boarding user interaction with ARIA
Prototype
Build Figma prototype to demonstrate the flow and the output
USER PERSONA
Name: Srikanth
Age: 32 years
Occupation: Working professional
I want to open joint account with my business partner
Generative AI can transform the way customer interacts with the bank and can emerge as is an intelligent finance partner providing an omni channel experience. Below is customer journey for scenario "opening joint account"
EXPLORING GENERATIVE AI
Designing Conversations
Creating a conversational AI for generative UI involves a careful consideration of user experience (UX) principles to ensure that interactions are seamless, intuitive, with focus on user. Our UX approach included:
​
-
Data Collection: A detailed study of possible use cases and Identify the specific contexts in which users will interact with the generative UI to tailor the conversation appropriately.
-
Persona and tone: Defined a consistent tone for the conversational AI that aligns with purpose of the idea. Ensured that the AI-generated content is relevant, helpful, and adds value to the user.
-
Document scan: Implemented OCR to convert scanned text into machine-readable format and provide required summary to the user
-
Feedback and Validation: Provide immediate feedback to user inputs to acknowledge their actions and maintain engagement. Clearly communicate if user missed any information or if AI system needs further clarification.
Architecture
user sends input
AI Chat Responses
UI Chat Box
Inputs to AI
AI Response
Open AI API (LangChain)
UI Chat Box:
The front end user interface where the user interacts and provides inputs to be sent to the AI.
-
Built using Python Libraries such as Panel.
-
Panel provides various templates to show various elements such as an input box and file upload features. Panel also provides the Text To Voice feature that is being used for the AI to talk.
-
Jupyter was the IDE used to help program ARIA. Jupyter provided us an easy way to interact with the Python libraries and compile code.
OpenAI API:
The open source AI API, specifically using LangChain, is used to provide basic chat features and also document readings.
We are taking advantage of the AI to read a document and return information to the user about the document.
CONCEPT VIDEO
