Vasooli AI.
Get paid on time, without the chase
Payment recovery for Indian MSMEs — reminders, statutory interest and escalation, on top of Tally, Excel and WhatsApp.
- Client
- Promptly AI
- Industry
- FinTech / MSME
- Engagement
- Build — four audited phases
- Relationship
- A Promptly AI product, built by the studio

The brief.
Small manufacturers across India lose weeks chasing unpaid invoices, and many never claim the interest the law already gives them on late payment. Vasooli AI set out to do the chasing for them — politely, in the buyer's language, on the channels they already use and on top of the books they already keep in Tally or Excel — while the owner decides what goes out.
Our response.
We built a NestJS API on PostgreSQL with row-level security, a responsive React web app in Hindi and English, a read-only Windows agent in Tauri and Rust that syncs TallyPrime and Excel, and a public website with GSAP and Lenis motion. The platform works out MSMED interest deterministically with the full working shown, sends reminders on WhatsApp, email and SMS through a tone ladder with approval modes, quiet hours and a send-time check, detects payments from syncs, bank statements and Razorpay, and reads buyer replies with AI. When patience runs out it prepares escalation: a demand notice drafted from ledger facts, an interest statement and an evidence bundle. A chartered accountant can follow every client business in one portfolio.
- Product Design
- Full-Stack Web Development
- Desktop Agent (Tauri)
- AI Integration
- Website & Motion Design
- Security & Privacy
Brick by brick.
- Course 01
Books First
Reads the ledger from Tally through a read-only desktop agent, from Excel or CSV, from reports sent by email or WhatsApp, or from bill photos confirmed by a person.
- Course 02
Interest You Can Prove
A deterministic MSMED interest engine with compound monthly rests and sourced RBI Bank Rate history, always shown with its full working.
- Course 03
Polite, Then Firm
A reminder engine with a tone ladder, approval modes, quiet hours, caps and a send gate that checks the books again at the moment of sending.
- Course 04
The Owner Decides
One activity feed, a daily WhatsApp digest with one-tap approval, a hash-chained audit log, and tenant isolation with Postgres row-level security.

What it became.
- API Routes
- 193
- Each with a declared access policy
- Automated Tests
- 733
- Shared, API, web, Rust agent and end-to-end suites
- Languages
- 2
- Hindi and English, across the app, website and agent
- Audits
- 4
- Independent audits, each closed with fixes and tests
Project scope and implementation details from Edifio’s case study.
Schedule of materials.
- 01NestJS
- 02React 19
- 03Vite
- 04Tailwind CSS 4
- 05Prisma
- 06PostgreSQL (RLS)
- 07Redis + BullMQ
- 08Tauri 2
- 09Rust
- 10GSAP
- 11Lenis
- 12WhatsApp Cloud API
- 13AWS SES
- 14MSG91
- 15Razorpay
- 16OpenAI / Anthropic
- 17Gotenberg
- 18Docker Compose







