How to Build an AI System That Routes Incoming Documents to the Right Department and Saves Your Admin Team 12 Hours Weekly

We built a document routing system for a mid-sized operations team using Make, Claude, and Gmail. It now sorts incoming invoices, contracts, HR forms, and compliance docs automatically. This guide covers the tools, the exact setup steps, and the honest limitations you need to know before you build.

What Is a Document Automation Workflow and Why Does It Matter?

A document automation workflow is a system that reads incoming documents, identifies what type they are, and sends them to the right person or folder without anyone touching them manually.

Right now, someone on your admin team is opening emails, reading attachments, and forwarding them by hand. That takes 2 to 3 minutes per document. At 200 documents a week, that is 6 to 10 hours gone. Add filing, logging, and follow-up, and you are easily at 12 hours weekly.

This system changes that. Documents come in. AI reads them. They route themselves. Your team handles exceptions, not every single file.

If you want to understand how this fits into a broader document strategy, How to Read Your Company's Processes Like an AI Person and Spot 20 Hours of Hidden Automation Opportunities in Your Department is a strong starting point.

Which Tools Should You Use?

You need three things: a trigger that catches incoming documents, an AI that classifies them, and an automation layer that routes them.

Here are the tools we tested and recommend.

We use Claude for classification. It handles long documents and ambiguous content better than most alternatives. ChatGPT and Gemini work too, but Claude's context window makes it more reliable when documents run long.

For the automation layer, Make costs less than Zapier at scale. Zapier is easier to set up if you have never used either. Both connect to Gmail, Outlook, Google Drive, and SharePoint without code.

If your team already uses PandaDoc for contracts, How to Build a Contract Review System Using PandaDoc and AI That Flags Compliance Issues Before Legal Sees Them shows how to layer AI review on top of routing.

How to Get Started Step by Step

• Map your document types. List every category your team receives: invoices, contracts, HR forms, compliance docs, vendor agreements. Aim for 5 to 8 categories.
• Create destination folders. In Google Drive or SharePoint, build one folder per category. Name them clearly: Finance-Invoices, Legal-Contracts, HR-Forms.
• Set up your Make account. Go to make.com and create a free account. Click Create a New Scenario.
• Add a Gmail or Outlook trigger. In Make, search for Gmail, select Watch Emails, connect your account, and set it to watch your shared inbox.
• Add a Claude API step. In Make, add an HTTP module. Point it to the Claude API endpoint at api.anthropic.com/v1/messages. Paste this prompt: "Read the attached document text. Classify it as one of these types: Invoice, Contract, HR Form, Compliance, Vendor Agreement, Other. Return only the category name."
• Add a router module. In Make, add a Router. Create one path per document type. Each path checks if Claude's output matches that category.
• Connect each path to a folder. Add a Google Drive or SharePoint module to each path. Set it to upload the file to the matching folder.
• Test with 10 real documents. Run the scenario manually. Check accuracy. Adjust your prompt if Claude misclassifies anything.
• Turn on scheduling. Set Make to run every 15 minutes. Your system is live.

Total build time: 90 minutes to 2 hours for a first-time setup.

What to Watch Out For

Scanned PDFs without text layers will break this system. Claude reads text, not images. If your team receives scanned documents, you need an OCR step first. Adobe Acrobat, Google Drive's built-in OCR, or a tool like Docparser ($39/month) can handle this before the document reaches Claude.

Also, Claude's classification is only as good as your prompt. Vague categories produce vague results. If you have documents that overlap, like a vendor contract that also includes an invoice, you need to decide your priority rule upfront. We recommend routing by the primary purpose of the document and flagging edge cases to a human review folder.

---

Someone in your department built this system last week. They are already recovering 12 hours every week while their admin team focuses on work that actually requires judgment. While you read this, the gap between you and them gets wider. Every week you wait is another 12 hours your team spends on manual sorting. Zero Day AI gives you mission files that tell your AI exactly what to build. You paste. It builds. You walk away with a working system in under an hour. Try it for $1. Two weeks. Full access. If it is not for you, cancel. But the gap does not close itself.

What to Do Right Now

Open Make and create a free account. That is the one action. Do not plan the whole system today. Just get the account open and connect your Gmail or Outlook inbox. That first step takes 8 minutes and makes everything else real.

If you want to see how this connects to a larger document strategy, How to Set Up AI to Auto Fill Client Documents From Your Email and Save 8 Hours Weekly Without Code shows the next layer you can build once routing is running.

Every week you wait, your admin team spends another 12 hours on work a $9/month tool could handle. Build it this week.