CreditPath AI

AI agent that builds a personalized 90-day credit roadmap for underbanked users — helping them become loan-eligible through automated guidance, weekly nudges, and financial scoring

Comment

The Problem We're Solving

Over 190 million adults in India are "credit invisible" — they earn money, pay bills, and manage households, but when they walk into a bank, they're turned away. Not because they're irresponsible. Because they've never had a credit product before. It's a catch-22: you need credit history to get credit.

We built CreditPath AI to break that cycle.

What Inspired Us

Living in India, we've seen this firsthand — a small shopkeeper denied a business loan, a young professional rejected for their first credit card, a family unable to finance their child's education. The information gap is massive. People don't know what lenders look for, what to fix, or where to start. We wanted to fix that with AI.

What CreditPath AI Does

CreditPath AI is an AI agent — not just a chatbot — that runs a personalized credit-readiness program for each user:

  1. Financial Intake: The user answers ~10 questions about income, expenses, existing loans, and banking habits
  2. AI Analysis: Gemini 1.5 Pro analyzes their profile and identifies the exact gaps keeping them from being credit-eligible
  3. 90-Day Roadmap: The agent generates a step-by-step, prioritized action plan tailored to that specific user
  4. Automated Weekly Nudges: Every week, an AI agent sends the user their next actions via WhatsApp — no human needed
  5. Credit Readiness Score: A simple 0–100 score that updates as users complete actions, showing real progress

The entire operation — intake to analysis to follow-up — runs with AI. A user can sign up at midnight and have a full roadmap in their inbox within 3 minutes.

How We Built It

  • Gemini 1.5 Pro (Google Cloud) powers all financial analysis and roadmap generation
  • Next.js for the frontend intake form and dashboard
  • Google Cloud Functions for the automated weekly nudge agent
  • Firebase for real-time database and user management
  • Razorpay for payments
  • Twilio for WhatsApp message delivery

The core flow is simple: form → Gemini API call with a structured prompt → parsed JSON roadmap → stored in Firebase → Cloud Function scheduler sends weekly updates.

Challenges We Faced

Prompt engineering for financial advice: Getting Gemini to produce actionable, safe, jurisdiction-aware credit guidance (not generic tips) took significant iteration. We landed on a structured output format with explicit constraints.

Keeping it AI-native, not AI-assisted: The temptation was to manually review every roadmap. We resisted. Instead, we improved the prompt until the output was consistently reliable — because the business only scales if AI runs it.

Getting real paying users on weekends: Cold outreach is hard. We focused on hyperlocal communities — college WhatsApp groups, local entrepreneur forums — and offered the first 10 users a discounted rate in exchange for honest feedback.

What We Learned

AI agents can genuinely replace entire workflows — not just assist them. The insight that changed everything: write your operations as prompts, not as code. The business logic lives in the prompt, not the backend.

We also learned that "real revenue" is the best forcing function. Knowing someone paid ₹499 made us take quality seriously in a way no hackathon deadline ever could.

Built With

Share this project:

Updates