TallyMsg
An AI personal finance SaaS where expenses are logged by sending a plain message on WhatsApp or Telegram, parsed automatically with no app to download.
The problem
Personal finance apps ask users to change behaviour before they get any value: download an app, create an account, categorise manually, and check back daily. Most people abandon that loop within a week. We wanted the entry point to be a habit people already had, sending a text, not a new one they had to build.
Our approach
We treated WhatsApp and Telegram as the interface rather than a notification channel. A user sends a message like "lunch 45 dirhams," and an LLM extracts amount, merchant, and category from unstructured text with no rigid command syntax to learn. The web dashboard exists for the parts messaging can't do well: charts, budgets, and trend lines, not data entry.
What we built
A Next.js dashboard and Node.js backend on Supabase, with webhook ingestion from the WhatsApp Business API and Telegram Bot API feeding an OpenAI-based parsing layer. Tiered billing unlocks voice note logging, receipt scanning, and AI-generated spending digests on top of the core message-based logging. We owned product scope, frontend, backend, the AI parsing layer, billing, and deployment as a single build.
The outcome
Expense logging dropped to the time it takes to send a text. Users on higher tiers can forward a receipt photo or a voice note instead of typing, and the AI digest surfaces spending patterns without anyone opening a spreadsheet. The product runs as a self-serve SaaS with no manual onboarding step required.