Match Tolerance & Auto-Approval Rule Manager
Define, simulate, and govern your invoice-matching tolerances and auto-approval rules - then see exactly how many invoices would auto-pass versus route to a human before you publish a single change.
A guarded internal tool where AP edits tolerance rules safely, simulates auto-pass vs exception rates and dollar exposure against real invoices, and a controller approves and publishes a versioned ruleset that the match workbench reads as the single live source of truth.
Before you start
- A free Supabase account
- A free Vercel account
- A free Resend account (for publish/alert emails)
- Your current tolerance settings and a recent sample of matched invoices (a CSV or Google Sheet is fine - no API needed)
The problem this kills
Every AP team runs on tolerances: price within 2% or $25, quantity within 5%, small-dollar invoices that should just clear. Those numbers decide what flows straight through and what lands in a human's queue. But here is the painful part - nobody actually knows what happens when you change them.
Widen the price tolerance to clear more low-touch invoices and you might quietly start auto-approving five-figure overcharges. Tighten it to be safe and you bury your team under exceptions that don't matter. The rules usually live buried in an ERP config screen or a controller's head, with no history of who changed what, when, or why. So changes either never happen (because they're scary) or happen blind (because someone guessed).
You shouldn't have to gamble. You should be able to type in a proposed tolerance, point it at last month's invoices, and see in plain numbers: "this would auto-pass 1,240 invoices, route 86 to a human, and expose $310,000 in auto-cleared dollars - including 3 invoices over $10,000." Then a controller looks at that, approves it, and it goes live as a versioned rule - not before.
What you'll build
A small, secure web app for your AP and controller team:
- Import your current tolerance settings and a recent set of matched invoices (from a CSV or Google Sheet - no integration required).
- Edit tolerances and per-supplier / per-category overrides in a guarded UI, as a draft that never touches live matching.
- Simulate the draft against your real invoices: how many auto-pass, how many route to a human, the total dollar exposure that would clear untouched, and a loud warning when a rule would auto-pass high-dollar items.
- Approve & publish: a controller reviews the simulated impact and publishes a new, numbered tolerance version with an effective date. Draft rules never affect anything until that moment.
- Read the active version cleanly - your match workbench (or your ERP export) always pulls the one published ruleset, with full version history behind it.
What's inside the Implementation Plan
This isn't a generic template you have to bend to fit. The plan opens by interviewing you about your business - your current match process and who owns it, your ERP and spreadsheets, the real field names and supplier/category codes in your data, your typical and peak invoice volumes, your exact tolerance and approval rules, and the messy exceptions you live with. It reflects back a short tailored spec, you give a thumbs-up, and only then does it build a tool shaped to how you actually work.
Inside you'll get:
- The full discovery interview, written out question by question.
- A step-by-step build, each step ending in a ready-to-paste prompt for your AI coding agent.
- A data model for rules, overrides, versions, and imported invoice samples - tuned to your answers.
- The simulation engine that scores auto-pass vs exception and totals dollar exposure.
- The approval-and-publish gate with versioning and effective dates.
- A "No API yet?" fallback so you can build the whole thing today from a CSV and export published rules back to Sheets/CSV in the exact columns your system expects.
- A verification checklist so you know it actually works before you trust it.
The governance it includes (this is the point)
This tool decides what clears your AP without a human touch - so it is governed to the teeth:
- Login so only your team can open it.
- Row-level security so each organization only ever sees its own rules and invoice data.
- A complete audit trail: every draft edit, every simulation run, every publish - who, what, and when.
- A hard human-in-the-loop gate: AP drafts and simulates, a controller reviews the impact and approves, and only then is a new tolerance version published and live. Drafts never affect matching.
- Duplicate guards: each ruleset has a signature (a hash of its conditions) and each publish gets a sequential version number, so the same rules can't be published twice and you always know which version is active.
- High-dollar warnings: the simulator flags any rule that would auto-pass large invoices, so a widened tolerance can't quietly wave through a big overcharge.
Who it's for
AP managers and controllers who want to safely widen or tighten what clears without a manual touch - and prove the impact before going live. If you've ever wanted to adjust a tolerance but didn't dare because you couldn't see the blast radius, this is your tool. No coding background needed.
You've got this - paste the first prompt and let the agent interview you.