Step 1: Pick your trigger

Start by deciding where your documents come in. Common triggers include:

• A new email with an attachment arrives in Gmail or Outlook

• A file is uploaded to a specific Google Drive or Dropbox folder

• A form submission comes through with attached files

• A message with a document is posted in Slack

Step 2: Use AI to classify the document

Add an AI step in your Relay.app workflow that reads the document and determines what type it is. Prompt Relay.app with something like:

"Read this document and classify it as one of the following: Invoice, Contract, Receipt, Onboarding Form, or Other. Also extract the date, company name, and any relevant reference number."

Relay.app gives you access to top AI models from OpenAI, Anthropic, Google, and more using built in AI credits, and it picks the best one for your use case. No managing API keys, no juggling separate token bills, no needing to know which model is good at what.

Step 3: Route and rename the file

Use the AI output to decide where the file should go and what it should be called. Add a Google Drive, Dropbox, or OneDrive step that uploads the file into the right folder using a clean naming convention like 2026-05-01_Acme_Invoice_INV-1042.pdf. Relay.app's branching logic makes this easy: One path for invoices, another for contracts, another for receipts.

Step 4: Log it for easy retrieval or analysis later

Add a step to log the document details in a spreadsheet, Notion database, or Relay.app's own Tables feature. This gives you a searchable index of every document that's been processed. Include fields like document type, date, sender, file link, and any extracted amounts or reference numbers.

Step 5: Add a human in the loop for edge cases

Not every document is clear cut. Sometimes the AI isn't sure what category something falls into, or the stakes are high enough that you want a person to confirm the classification or extracted data before filing.

This is where Relay.app's human in the loop feature comes in. You can add a checkpoint that pauses the workflow and sends you or a teammate a notification to review the classification before the document gets filed. Human-in-the-loop checkpoints are a safety net when accuracy matters more than speed.

Once you approve, the workflow picks back up and files everything correctly. You get the best of both worlds: automation for the routine stuff, human judgment for the tricky stuff.

Tips to make your document filing workflow even better

• Start simple, then expand. Get the basic classification and routing working first. You can always add more document types and smarter logic later.

• Use Relay.app Tables to track everything. A central log of processed documents makes auditing and searching easy, and gives you a foundation to build more automations or analysis on top of later.

• Set up separate workflows per channel. One for email attachments, one for Slack uploads, one for form submissions. Each flow stays simple, easy to debug, and easy for teammates to understand at a glance if you're building shared flows.

• Test with real documents first. Use Relay.app's free test runs to validate your workflow with actual files before going live.

• Collaborate with your team. Share workflows using Relay.app's collaborative features so teammates can refine prompts and add their own filing rules.

• Use the Relay.app Assistant to build, test, and improve. You don't have to build this all alone. Just chat with the AI assistant to explain what you want. It builds the workflow for you and helps you test and ship it.