On-Chain AI Treasury Assistant

Inspiration

Working in fintech, I saw how companies handle treasury management manually. A Citibank employee once caused a $900 million loss by fat-fingering a loan repayment, sending the full principal instead of just interest. This got me thinking about how many smaller mistakes happen daily in corporate treasuries that nobody talks about. Small businesses face the same problem on a smaller scale. CFOs manually move money between accounts, often keeping too much cash idle because they're busy with other priorities. The result is thousands in lost yield every year.

What it does

Most small businesses earn 0.1% on idle cash while safe options offer 4-5%. The manual process means:

Money sits idle for weeks Opportunities get missed Human errors happen during transfers No systematic approach to risk management

How we built it

Started with synthetic treasury data to simulate a real company with multiple departments. Built the AI recommendation engine using basic portfolio optimization principles. The system analyzes idle funds, risk tolerance, and available opportunities to suggest moves between cash, stable yields, and crypto allocations. Used Next.js for the frontend to focus on the user experience. The AI logic simulates what would connect to OpenAI's API in production. For crypto allocations, integrated Sui wallet functionality to execute blockchain transactions when the AI recommends moving funds to digital assets based on the company's risk profile. The Sui integration handles the actual movement of funds to crypto when recommended, with smart contracts ensuring secure execution of AI-driven portfolio rebalancing decisions.

Challenges we ran into

The biggest challenge was balancing automation with control. Treasury managers need to trust the system but also maintain oversight. I solved this by making every recommendation transparent with clear risk levels and expected returns. Another challenge was portfolio allocation logic. Different companies need different risk profiles. A restaurant needs liquid reserves, while a tech startup might want crypto exposure. The system now adjusts recommendations based on business type and cash flow patterns. Integrating Sui wallet functionality added complexity around transaction security and gas optimization, but the low-cost blockchain makes it practical for frequent small rebalancing operations.

What we learned

Building this taught me that AI doesn't need to be complex to be valuable. Simple rule-based logic combined with portfolio theory can make better decisions than busy humans. The key insight was making the system suggest actions rather than execute them automatically, keeping humans in control.

What's next for On-Chain AI Treasury Assistant

The prototype proves the concept works for basic treasury optimization. Next steps include connecting to real banking APIs, adding actual compliance checking, and building risk models for different industry types. The goal is to help small businesses stop losing money on manual treasury management while keeping humans in control of their financial decisions.