Driftrecon

Inspiration

Multi-platform businesses struggle to reconcile sales, fees, refunds, FX, disputes, payouts, and bank deposits when they happen across different systems and accounting periods. We built DriftRecon to make that process automatic and explainable.

What it does

DriftRecon connects financial events from sale to bank deposit, automatically reconciles safe matches, investigates ambiguous cases with an AI agent, and sends uncertain cases to human review with evidence.

How we built it

We used Next.js and TypeScript for the app and reconciliation engine, TensorMux for the Recon Agent, Neatlogs for agent tracing, Dodo Payments as a payment source, and Agent Orchestrator to coordinate development and testing.

Challenges we ran into

The main challenge was handling cross-period refunds, FX, bundled payouts, duplicate amounts, and ambiguous matches without letting the AI control financial calculations.

Accomplishments that we're proud of

We built an end-to-end reconciliation flow where deterministic code handles financial truth, the agent investigates exceptions, humans approve uncertain cases, and approved decisions become reusable reconciliation policies.

What we learned

We learned that reliable financial automation needs clear separation between deterministic accounting logic, AI reasoning, and human judgment.

What's next for DriftRecon

We want to add more payment processors, direct bank integrations, accounting-system integrations, continuous reconciliation, and a larger policy library built from approved finance decisions.

Built With

• agent
• orchestrato
• typescript