Denied-Party Screening Log: Catch Restricted Buyers Before You Ship
Screen every customer and consignee against an imported denied-party watchlist with fuzzy matching, then have compliance clear, hold, or escalate each hit — before the order moves.
A web tool where you import your orders and a denied-party watchlist, AI fuzzy-matches every party name and address against the list and explains each potential hit, compliance reviews and approves a clear/hold/escalate disposition, and it exports cleared orders plus a full screening audit log — with periodic re-screening built in.
Before you start
- A Supabase account (free)
- A Vercel account (free)
- A Resend account (free)
- A CSV of your parties/orders to screen and a CSV denied-party watchlist
- Claude Code or any AI coding agent
The problem this kills
Somewhere between an order coming in and a truck leaving the dock, someone is supposed to make sure you're not shipping to a sanctioned company, a denied person, or a restricted address. Miss it, and you're not looking at a clawback — you're looking at fines, seized shipments, and an export-control violation that can follow the company for years.
The maddening part is that the screening itself is mechanical: take the buyer's name and address, compare it against the denied-party lists, and decide whether anything that lights up is a real match or just a coincidence. But names are messy. "Acme Trading Co." on your order is "ACME TRADING COMPANY LTD" on the list. A consignee in one country shares a name with a sanctioned entity in another. A single transposed letter hides a true hit, and a dozen near-misses bury your reviewer in noise. You need something that screens every party against the list, surfaces the potential matches with a plain-English reason, and then puts a trained compliance person in charge of the call. You do not need to be a developer to build that something.
What you'll build
A simple internal web tool for trade compliance and order management. You import two files: the parties/orders you need to screen (customer and consignee names plus addresses), and a denied-party watchlist as a CSV — because there's no live feed, the file is the list. The tool fuzzy-matches every party against the watchlist, catching legal-suffix differences, reordered words, transliteration variants, and partial address hits, and attaches a match score and a plain-English reason to each potential hit. Your compliance reviewer opens the queue and, for every flagged party, records a disposition: Clear, Hold, or Escalate — with a reason. Only cleared orders are released: the tool exports a cleared-orders CSV in your system's exact columns plus a complete screening log of every hit and decision. And because watchlists change, it can re-screen open orders whenever you import an updated list.
What's inside the Implementation Plan
The downloadable plan is a step-by-step file you paste into an AI coding agent. It opens by interviewing you about your business — your current screening process and who owns it, the order/ERP system your party data comes from, the exact fields and naming conventions in your records, your typical and peak screening volumes, your real clear/hold/escalate rules and thresholds, and the messy edge cases (transliterated names, freight forwarders, intermediate consignees, common names). It reads a short spec back to you for a thumbs-up, then builds the tool around your answers instead of a generic template. From there it walks the agent through the data model, the two CSV imports, the fuzzy-matching engine and how to tune it, the review-and-decide screen, the human disposition gate, periodic re-screening, and the cleared-orders and screening-log exports. Every step ends with a ready-to-copy prompt.
The governance it includes (this is the point)
This isn't a toy. The plan builds in the controls a real compliance function needs: login so only your team can use it, row-level security so people only see their own organization's orders and watchlist, a complete audit trail of every disposition and override (who, what, when, and why), a hard human-approval gate so no order is released to ship until a person has cleared every potential hit, and duplicate guards so the same party-and-order can't be screened or processed twice. The whole tool exists to make a careful human decision easy and defensible — the AI raises the suspicion and shows its work; a trained person makes the call, and the record proves it.
Who it's for
Trade-compliance officers, export-control leads, and order-management teams who own the "are we allowed to ship to this party?" question and need a defensible record of every screening decision. If you can describe what makes a name "a match" in your world, you can build this.
You've got this — open the plan, paste the first prompt, and you'll be screening your first real batch against a watchlist this weekend.