跳到正文
DEV Community· Tirumala Rao·· 8 天前AI 評分56

FinMoment 建造者分享 AI 記帳流程的 5 項設計規則

AI shouldn't guess with money: 5 design rules from building an AI bookkeeping pipeline

AI 導讀

FinMoment 的建造者分享 5 項 AI 記帳流程設計規則,主張由模型讀取文件,再由程式按固定規則驗證結果,並把不確定欄位交由人員審核。流程會核對每頁結單的期初餘額、貸項、借項與期末餘額,並將每筆交易連回原始明細行。FinMoment 設有 MCP 端點供 Claude 操作,但最終核准由人員負責;作者亦提到團隊使用 13 個 Claude Code 智能體,並設有 CI、部署及核准關卡。

正文

Most AI demos for finance look great until the numbers have to add up. A model that is 97% right on bank statements is still wrong on three lines out of a hundred, and in bookkeeping one wrong line means the books don't close.

I spent about 15 years building large systems, most recently as a solution architect and engineering manager on fraud-decisioning platforms at Capital One. When I started building FinMoment, an AI bookkeeping pipeline, I brought one rule from that world with me: the system is allowed to be unsure, but it is never allowed to guess.

Here are the five design rules that came out of it. They apply to any AI agent that touches money, invoices or documents people rely on.

1. Let the model read, but let code decide

The pipeline is: upload, then OCR, then extraction, then a review queue, then deterministic validation, then human approval, then export (CSV, Excel or Tally XML).

The language model does the parts it is good at: reading messy statements and receipts, and turning them into structured fields. It does not decide whether the result is correct. That is done by plain code with rules that can't be talked out of an answer.

2. Make the numbers prove themselves

The most useful check is also the simplest: balances must reconcile. Opening balance plus credits minus debits has to equal the closing balance, for every page of every statement. If it doesn't, something was misread, skipped or duplicated, and the run stops there instead of producing a nice-looking file that's wrong.

A check like this catches errors a confidence score never will, because the model can be very confident and still be wrong.

3. Send uncertainty to a human, field by field

When a field is unclear (a smudged amount, an ambiguous date, a merchant name the model can't place), it goes to a review queue instead of being filled in with a best guess. The person sees exactly which fields need attention, not a whole document to re-check.

This is the difference between "AI that saves time" and "AI that creates a second job of checking the AI".

4. Link every number back to its source

Every transaction is labelled and linked to the line it came from. When an accountant asks "where did this come from?", the answer is one click away. Without that, nobody can audit the output, and output nobody can audit is output nobody can trust.

5. Agents can drive the system, but humans approve

FinMoment has an MCP endpoint, so Claude can drive the pipeline. The final approval step is human-only. I run the product with a team of 13 Claude Code agents, and the same pattern applies there: done-checks, CI and deploy gates, and approval gates that only a human can pass.

Agents are fast and tireless. They are also confidently wrong sometimes. Gates are how you keep the speed without inheriting the mistakes.

What this means if you're adding AI to a finance or document workflow

Before picking a model, write down three things:

  • What check proves the output is right? (Totals, reconciliations, cross-field rules.)
  • Where does uncertain data go? (A queue with the exact fields, not a full re-review.)
  • Who approves, and what do they see? (The source, the extracted value and the reason it was flagged.)

If you can answer those, the model choice matters much less than people think.

I help teams build this kind of AI automation: document and invoice extraction, bookkeeping workflows, and Claude agents with approval gates and audit trails, usually on AWS or a Next.js/Supabase stack. If you're working on something similar, I'm happy to walk you through FinMoment on a call. You can find me on Upwork: https://www.upwork.com/freelancers/~0146d40003db2cffd3

Originally published on LinkedIn.

來源:DEV Community · dev.to