An Intelligent Business Analytics Chatbot for SME Financial Intelligence Using Hybrid AI Architecture

Abstract

For this project, I built and tested a chatbot called Business AI Consultant. It’s designed for small and medium-sized businesses—exactly the kind of companies that don’t have huge budgets for fancy analytics. The whole idea is to make really smart, data-driven business advice as easy to access as possible.

Here’s how I did it: I combined a Retrieval Augmented Generation (RAG) setup so the bot can grab the right info, with a finance-focused engine built on LangChain. Even non-technical business owners can use a simple Streamlit web page to upload their CSV files, ask questions, and get complex insights back—things like profit breakdowns, trend spotting, periodic summaries, or strategy suggestions. Under the hood, it’s powered by GPT-4o-mini and a custom ReAct prompt. That way, the system keeps conversations lively but still gets the math right.

From initial testing, I’ve seen this system turn raw transaction logs into actually useful business intelligence. It proves you don’t need an endless budget to get advanced analytics—tools like this can really level the playing field. I’ll finish up by going over some of my design choices, the challenges I ran into, and ideas on how to make the platform even more robust and insightful.