Helpers on call
Retail journey
First salary, first system.
Priya, 24 · QA engineer in Coimbatore · first job
Priya gets her first salary on the 30th. She wants a system that holds her plan together — not a spreadsheet she will abandon by month two.
Marketing claimeight-languages
The story · 5 scenes
Beginning to end.
Priya, 24 · QA engineer in Coimbatore · first job
·Priya gets her first salary on the 30th. She wants a system that holds her plan together — not a spreadsheet she will abandon by month two.- Scene 01 · Sign in once, in TamilPriya enters her phone number, types a 6-digit OTP, and picks Tamil as her language. The app reads everything back to her in Tamil from the next screen.
- Scene 02 · Connect her salary accountShe picks her bank from the Account Aggregator bottom sheet, consents on her bank's screen, and watches her last six months of transactions land — categorised, tagged, and totalled.
- Scene 03 · See her own health scoreBy that evening, Mukodi has computed a financial health score from her income, EMI, and savings rate. The reasons are written in plain Tamil; the trend line shows the next three months.
- Scene 04 · Set a first goalPriya sets a six-month emergency fund target. The app shows a P10–P90 forecast band instead of a fragile single line — and tells her, gently, which month she is most likely to fall behind.
- Scene 05 · Ask Mukodi a questionShe asks, in Tamil, whether she should pre-pay her education loan or build her emergency fund first. The answer comes back with three reasons, each linked to her own transactions.
Helpers on call
The stack · helpers + models
Who's actually doing the work.
Models invoked
Honest limits
What this flow doesn't do.
- The AA connection requires Priya's bank to be live on the AA network — most major Indian banks are, but a few co-operative banks are still in pilot.
- Health-score reasons are operator-vetted templates; the underlying model output is converted to plain language and is not a financial-advice instrument.
- Voice replies use IndicTTS pronunciations; very long numerical strings may be read digit-by-digit when the locale-specific TTS pack lacks a fluent decoder.