Inspiration
Growing up in Nigeria, I watched my mother coordinate an ajo group every month. Ten women, each contributing ₦50,000, taking turns collecting the full pot. It ran entirely on trust, WhatsApp messages, and a notebook.
One month, a member swore she had paid. My mum - the coordinator said she hadn't. What followed wasn't just a financial dispute, it was a friendship breaking apart over ₦50,000.
That moment stayed with me. Over 10 million Nigerians participate in rotating savings groups called ajo (also known as susu in Ghana, chit funds in India, tontines in West Africa). Hundreds of millions of people worldwide use this centuries-old system. And every single group is managed the same way: WhatsApp, paper, and hope.
I built Ajo AI to fix that.
What it does
Ajo AI is a full-stack digital manager for rotating savings groups. It brings trust, transparency, and intelligence to a system that has operated on faith alone for centuries.
For Coordinators:
- Create a savings group with contribution amount, frequency, and rotation order
- Invite members via a shareable link
- Record payments as they come in, member statuses update in real time
- Mark payouts as disbursed when each round closes
- Use the AI Dispute Mediator when conflicts arise
- Send AI-generated personalised payment reminders to late members
For Members:
- See your payment status, when you collect, and your full history
- Know exactly when your payout round arrives and how much you'll receive
- View your Trust Score — built from your payment consistency over time
The AI Dispute Mediator is the crown jewel. When a coordinator describes a conflict, the AI analyzes the group's full payment history and rotation record, then returns: a situation summary, a fair data-driven resolution, and a ready-to-send message to both parties. No more arguments. No more broken friendships.
How we built it
Ajo AI was built entirely using MeDo — Baidu's AI-powered full-stack app builder.
I structured the entire build as a series of precise conversational prompts to MeDo, each building on the last:
Foundation — I described the app concept, user roles, and design aesthetic (deep green #1B4332 + gold #D4A017). MeDo generated the landing page, authentication system, and dashboard shell.
Database architecture — I specified the full schema (users, groups, memberships, rounds, contributions, disputes, reminders) and MeDo generated the backend and database with proper relationships and RLS policies.
Core features — Multi-step group creation, member invitation via unique links, coordinator-only payment recording, automatic round progression when all members pay.
AI integrations — I wrote system prompts for both the Dispute Mediator and Smart Reminder generator. MeDo wired these into functional chat interfaces connected to an AI API.
Polish — Dashboard stats wired to real data, payout disbursement confirmation flow, member trust score calculation, round status display logic, mobile responsiveness.
MeDo handled the React frontend, Node.js backend, PostgreSQL database, and API integrations automatically. I focused entirely on product thinking, user flows, and the cultural context that makes this solution meaningful.
Challenges we ran into
Infinite recursion in RLS policies — MeDo's generated Row Level Security policy on the group_members table referenced itself, causing a database error on group creation. Fixed by rewriting the policy to use direct auth.uid() checks.
Round recipient assignment — The initial rotation logic generated rounds with null recipient_ids, meaning no member was assigned to collect any round. Fixed by ensuring recipient assignment runs immediately and atomically at group creation.
AI streaming vs JSON parsing — The AI API was returning SSE streaming responses but the code was trying to parse them as plain JSON. Fixed by switching all AI calls to non-streaming mode.
Round status display logic — The UI was calculating round status from date ranges alone, ignoring the database status and disbursed fields. Fixed by establishing a clear priority: disbursed > completed > in_progress > upcoming.
Accomplishments that we're proud of
- Built a complete, deployed, full-stack application in 10 days using only MeDo
- The AI Dispute Mediator works end-to-end: it ingests real group payment data and returns contextually aware, culturally sensitive resolutions
- Solved a problem that affects hundreds of millions of people globally with zero dedicated digital infrastructure
- The app correctly models the full ajo lifecycle: group creation → member invitation → contribution tracking → automatic round progression → payout disbursement → dispute resolution
What we learned
MeDo is remarkably capable when given precise, structured prompts with full context. The key insight was treating each conversation as a product specification — the more specific and culturally grounded the prompt, the better the output.
I also learned that the hardest part of building for informal financial systems isn't the technology — it's deeply understanding the human dynamics (trust, fairness, relationships) that the technology must support without breaking.
What's next for Ajo AI
Ajo AI has a clear roadmap toward becoming a real fintech product:
V2 — Digital payments: Integrate Paystack so members can pay their contributions directly on the platform. Funds held in escrow, released to the recipient when the round closes. The coordinator never touches cash.
V3 — Mobile app: A React Native app optimised for the low-bandwidth, mobile-first reality of most Nigerian users.
V4 — Credit layer: A member's ajo payment history across multiple groups becomes a verified credit profile. Consistent ajo participants unlock access to microloans — a genuinely new credit scoring mechanism for the unbanked.
The informal savings system has operated on trust for centuries. Ajo AI gives that trust a digital backbone.
Built With
- ai-api
- javascript
- medo
- node.js
- postgresql
- react
- rest
- supabase
Log in or sign up for Devpost to join the conversation.