Track 01: Money, Jobs & AI — Will AI Steal My Future?

Inspiration

Most financial apps solve the knowledge problem. They tell you where your money goes, but knowing was never the issue. I kept failing my own goals not because I lacked information, but because I lacked skin in the game and people watching. I was inspired by the behavioural science of loss aversion and social commitment contracts, and I wanted to build the accountability layer that every budgeting app is missing.

What it does

Kova turns your friend group into a financial accountability system. You set a financial goal, your group stakes a shared pool, and an AI quest engine generates a personalised weekly spending challenge based on your actual transaction data. Complete the quest and your stake unlocks as a voucher reward. Miss it and your stake remains locked in your Kova wallet for 30 days, still yours, but illiquid. Your group sees your streak, never your financial data. The sting is accountability, not shame.

How I built it

I built Kova as a React and TypeScript web app with Supabase (Postgres) as the backend and Claude AI powering the quest engine. The AI analyses spending categories, identifies the highest controllable discretionary leak, and generates a specific dollar-target challenge that gets harder as your streak grows. The wallet system simulates a partner bank sub-account model with lock, release, and credit instructions, without Kova ever holding user funds. Auth and onboarding are handled via Supabase Auth, with a mobile-first UI built in Tailwind CSS.

Challenges I ran into

Designing the ethical mechanic was harder than expected. Making loss aversion feel motivating rather than punishing required careful balance. I spent significant time tuning the AI prompt so quests felt personal and achievable, not generic. Simulating the wallet and bank sub-account logic in Supabase without a real financial integration, while keeping the demo credible, was another key challenge. I also had to carefully think through privacy, ensuring the group accountability layer never exposes sensitive financial data.

Accomplishments that I'm proud of

I am proud of the AI quest engine. It genuinely feels personal. The behavioural design, stacking loss aversion, social commitment, and pre-commitment, is grounded in real research and translated cleanly into product mechanics. I shipped a full working prototype with auth, onboarding, AI quest generation, a group screen, wallet simulation, and an insights dashboard, end to end, in a hackathon timeframe.

What I learned

I learned that the hardest part of fintech is not the finance, it is the behaviour. Designing for motivation, not just information, requires a completely different product mindset. I also learned how powerful AI-generated personalisation is for engagement: a quest that references your actual spending feels fundamentally different from a generic tip.

What's next for Kova

Real bank integration via Plaid and SGFinDex for live transaction data. A production wallet backed by a partner bank sub-account. Expanding beyond financial goals to savings milestones and investment habits. A mobile app via React Native. And a B2B play, offering Kova as an employee financial wellness tool for corporates.

Built With

Share this project:

Updates