LedgerPilot AI

LedgerPilot AI is an AI finance employee for lean cross-border teams. It monitors cash, invoices, and treasury, while keeping every payment under human approval.

Comment

One-line description

LedgerPilot monitors a company’s receivables, treasury, and cash flow, then turns financial events into a daily brief and safe, approval-ready actions.

The problem

Lean teams do not lack finance software; they lack the time and context to operate it. Invoices live in inboxes, payments live across bank accounts and wallets, and cash decisions are rebuilt in spreadsheets. Traditional tools record financial activity but still wait for a founder to notice the issue, interpret the numbers, and perform the task.

The solution

LedgerPilot is an AI Finance Employee called Alex. Every morning, Alex presents work already completed, exceptions that require attention, and decisions that need a human. A founder can ask questions such as “Can we hire another engineer?” or “Pay the team,” receive an evidence-backed answer, and review a prepared action before signing it in their wallet.

What makes it different

  • Proactive rather than passive: LedgerPilot finds work and prepares the next step instead of waiting for dashboard navigation.
  • Evidence before action: Recommendations expose sources, assumptions, confidence, and expected impact.
  • Human authority: The AI can recommend and prepare, but it cannot sign or move funds.
  • One operational story: Morning Brief, Finance Inbox, Conversation, and Approval form a continuous work loop rather than separate modules.

Initial customer

The MVP is designed for 5–30 person AI automation agencies in Southeast Asia that receive cross-border income, manage multiple currencies or stablecoins, and do not yet have a full finance team.

Stage 1 concept

The current design demonstrates four core moments:

  1. Alex opens the day with a Morning Brief.
  2. The Finance Inbox triages exceptions and completed work.
  3. AI Conversation turns business questions into evidence-backed decisions.
  4. Approval Flow makes every money-moving action legible and human-controlled.

Technical direction

  • Next.js 15 and TypeScript for the product and API orchestration.
  • Supabase for authentication, Postgres data, and evidence storage.
  • OpenAI Structured Outputs for classification, reasoning, alerts, and actions.
  • wagmi and a user-controlled wallet for approvals.
  • Base Sepolia and a focused TreasuryExecutor contract for proposed, approved, and executed payment plans.

The security boundary is deliberate: the model may prepare a structured action, but policy checks run before submission and only the user’s wallet can sign.

Interactive prototype delivered

The deployed prototype now demonstrates the complete judge-visible loop:

“Pay the team.” → Retrieve company facts → Explain the plan → Review recipients and policy checks → Approve in wallet → Confirm on Base Sepolia → Save the audit receipt.

The Stage 1 build uses deterministic AI responses and a simulated wallet/receipt so the complete experience is reliable and safe to judge. The repository documents the live prototype separately from the planned OpenAI, Supabase, wagmi, and Base Sepolia integrations.

Long-term vision

LedgerPilot starts with the daily finance operations of lean agencies and grows into the financial operating layer for internet-native companies: one accountable AI employee coordinating receivables, treasury, cash planning, and controlled execution.

Asset disclosure

All company names, balances, invoices, confidence values, recipient counts, and transaction details shown in the prototype are fictional demonstration data.

Built With

Share this project:

Updates

Submission history