Inspiration

I wanted to make managing payments easier using AI. Instead of checking dashboards and handling everything manually, I thought, why not build an assistant that can understand financial questions and help users take action?

That's how PayPilot AI started.

What it does

PayPilot AI is an AI-powered financial assistant that helps businesses:

  • Analyze transactions and get useful insights.
  • Ask financial questions in simple language.
  • Create refund requests using AI.
  • Approve or reject refunds before execution.
  • Process approved refunds through PayPal Sandbox.
  • Track refund requests and their statuses.

The dashboard currently uses sample transaction data.

How we built it

I built PayPilot AI using React, TypeScript, Java Spring Boot, Google Gemini, PayPal APIs, and H2 Database.

Gemini understands user requests and routes them to the right agent. The agents analyze financial data or prepare refund actions. Refunds require human approval before execution through PayPal.

Challenges we ran into

The biggest challenge was making AI-generated actions safe and reliable, especially when dealing with real payment APIs.

I also faced issues with PayPal Sandbox permissions, handling refund failures, preventing duplicate requests, and saving refund history.

Since I was building it alone, managing everything within the hackathon timeline was another challenge.

Accomplishments that we're proud of

I'm most proud of building a working AI assistant that does more than just answer questions.

I successfully integrated Gemini with PayPal Sandbox, tested a real $2 refund, and built a human approval system to keep financial actions under control.

Completing the frontend, backend, AI integration, and testing on my own made it even more rewarding.

What we learned

I learned that building AI agents isn't just about calling an LLM. It's about understanding user intent, planning actions, handling errors, and making sure those actions are safe.

I also gained practical experience with Gemini, PayPal APIs, agent workflows, and building an end-to-end AI application.

What's next for PayPilot AI

I want to connect real PayPal transaction data, add more agents for payment monitoring, disputes, and financial insights, and improve security.

My long-term goal is to turn PayPilot AI into a useful financial assistant that helps businesses understand their money and take action safely.

Built With

Share this project:

Updates

Submission history