AI‑Powered Financial Report Summarizer (Apr 2025)

Summary

This project aims to develop a personalized AI-powered tool for summarizing financial reports. The tool is designed to enhance the efficiency of financial data analysis for users.

The use case are the financial reports for 2024Q4, 2025Q1 and 2025Q2 for the 6 big banks in Canada:

• Royal Bank,
• Toronto-Dominion Bank,
• the Bank of Montreal, (Not available)
• the Canadian Imperial Bank of Commerce, (Not available)
• the Bank of Nova Scotia
• National Bank of Canada

The desired format of the output is a table summarizing the financial numbers reported during the quarter. As I want to focus on the most recent financial reports, but the LLMs models may not have training data with this information, I am using Retrieval Augmented Generation (RAG).

Built as capstone project for 5-Day Gen AI Intensive Course with Google.

Objective

To create an AI-powered tool that provides personalized summaries of financial reports, making complex financial data more accessible and actionable for users.

Business Problem:

Financial reports are often lengthy and complex, making it difficult for users to quickly extract relevant information. Key challenges include:

• Time-consuming analysis of extensive financial documents
• Difficulty in understanding complex financial terminology and data
• Need for personalized insights based on individual user preferences

Business Opportunity:

There is an opportunity to leverage AI to transform the way financial reports are analyzed and understood. The new tool can:

• Generate concise summaries of financial reports tailored to user preferences
• Provide accurate and reliable insights using advanced AI capabilities
• Integrate seamlessly with other financial tools and systems

GenAI Capabilities Used:

• Document understanding: To comprehend and process the content of financial reports
• Grounding: To ensure the accuracy and reliability of the information provided
• Embeddings: To capture semantic meaning and provide relevant insights
• Retrieval augmented generation (RAG): To fetch relevant information from external sources
• Structured output/JSON mode/controlled generation: To generate structured outputs in table format.