Finpilot

An AI financial advisor that helps business owners understand performance, simulate decisions, and turn business data into actionable insights.

Comment

Inspiration

Commerce teams often have plenty of data but still struggle to turn it into confident decisions. Important signals are scattered across sales, marketing, inventory, and product reports. I created FinPilot to bring that information together and transform it into clear, evidence-backed actions.

What it does

FinPilot is an AI decision workspace for commerce businesses. Users can upload their business data and:

  • Monitor revenue, profit, marketing efficiency, and other key metrics
  • Discover anomalies and business risks
  • Trace every insight back to its supporting data
  • Ask the AI Advisor questions about performance
  • Test decisions in Scenario Studio before acting
  • Review recommendations and their expected impact

For example, FinPilot can identify a decline in ROAS, connect it to campaign performance, highlight low-stock bestsellers, and simulate the effect of reducing advertising spend.

How I built it

FinPilot was built as a full-stack web application using Next.js, React, TypeScript, Convex, Clerk, and the OpenAI API. It is deployed on Vercel.

I built the project with Codex powered by GPT-5.6. Codex helped me implement the application, develop the analytics logic, improve the interface, diagnose bugs, run tests, and refine the overall product experience. I reviewed the generated work and made the final product and design decisions.

Challenges I ran into

The main challenge was turning separate datasets into trustworthy insights rather than generic AI summaries. I had to design validation, metric calculations, evidence links, and scenario logic so users could understand why each recommendation was made.

Another challenge was creating a polished experience that remained useful even when AI services were unavailable.

Accomplishments that I am proud of

  • Built a complete, working product rather than a static prototype
  • Created evidence-backed insights with source traceability
  • Added an AI Advisor grounded in workspace data
  • Built an interactive scenario simulator
  • Prepared synthetic test data that exposes realistic business problems
  • Made the product accessible through a deployed demo

What I learned

I learned how effectively Codex and GPT-5.6 can accelerate the full development cycle from architecture and implementation to testing, debugging, and product design while keeping the human responsible for direction and final decisions.

What's next for FinPilot

Next, I plan to add more commerce integrations, collaborative decision workflows, automated monitoring, and richer forecasting capabilities.

Built With

Share this project:

Updates