aSKii

Khata is an AI-powered financial assistant that I developed to help small Indian kirana shop owners identify financial losses hidden in their everyday WhatsApp conversations with distributors. Shop owners regularly receive invoices, product price lists, payment confirmations, promotional offers and voice messages through WhatsApp. Keeping track of all this information manually is difficult, and small mistakes can result in missed discounts, incorrect payments or duplicate bills. Khata aims to solve this problem by analysing exported WhatsApp conversations along with their attachments, including images, PDFs, text messages and voice notes. I integrated Meta's Muse models to understand these different types of inputs and extract useful information from them. The extracted information is converted into structured records and analysed using Python-based validation logic. Khata can identify issues such as product rate mismatches, duplicate bills, missing promotional schemes, payment mismatches, GST errors and credit-period risks. One of the main things I focused on while building Khata was reliability. Rather than simply reporting an issue, the system provides supporting evidence from the original messages or documents. When information is missing or conflicting, it flags the situation for the shop owner's review instead of making assumptions. It also generates ready-to-use WhatsApp replies to help shop owners communicate with distributors. My goal with Khata is to make AI useful for everyday businesses by helping small shop owners save time, avoid financial mistakes and identify money they might otherwise lose.