Case Study Details

AI Chatbot

Focus

Design and development of a proof of concept for an AI powered chatbot that referenced and retrieved Agency files and data.

Approach

The Agency had a large collection of scanned and electronically generated documents that was incredibly tedious to manually filter and analyse. The chatbot analysed and integrated with the Agency's extensive database, document repository, and FAQ material to provide a natural language interface to query and search relevant information within the documents and database.

The project team first mapped out the system diagram and use cases, then key information was extracted from the documents utilising OCR technology. This information was extracted, transformed and stored in a database detailing different data components and extractions for each document.

A Retrieval Augmented Generation approach was used to create database queries to access the Agency's relational database and return the relevant information, documents and links augmented into the natural language response of the chatbot.

The chatbot was designed, developed and deployed in the Azure platform, handling FAQs as well as custom questions that returned information from the database and OCR extraction as well as the relevant documents.

Outcome

• Reduced time for analysis and retrieval of information from hours to seconds
• Improved accuracy and timeliness of information
• Increased employee capacity for higher value activities