LaunchPath Inc. is developing a Minimum Viable Product (MVP) for an AI-powered interactive bookkeeping assistant. The goal is to create a system that can scan and extract information from bank statements, receipts, and invoices, automatically categorize transactions, and interact naturally with users to confirm uncertain entries.

Unlike static automation tools, this agent will learn user preferences through conversation — asking clarifying questions, remembering context, and improving accuracy over time. The project’s objective is to demonstrate how LLM-driven automation can simplify financial management for freelancers and small businesses while maintaining transparency and user control.

The main deliverable will be a functional MVP prototype of the AI bookkeeping assistant that can:

• Upload and parse financial documents (bank statements, receipts, invoices).
• Extract transaction data using OCR and basic NLP.
• Categorize transactions automatically with editable fields.
• Interact with users through a chat interface to confirm or fill in missing data.
• Learn from user corrections and improve categorization accuracy over time.
• Generate basic financial summaries and CSV/Excel exports.

A final project report will outline the system design, AI workflow, learning behavior, and testing results, along with future development recommendations.