Skip to main content

​
Goal

Your first workflow should teach you the product, not test your patience. Aim for a workflow that:
  • has one clear trigger
  • has one clear outcome
  • uses only a few steps
  • is easy to test
Examples:
  • receive a webhook and send a Slack message
  • run on a schedule and generate an AI summary
  • capture a form or email event and write a record into Google Sheets

​
Step 1: Start with the outcome

Before opening the editor, decide what success looks like. Good examples of clear outcomes:
  • a message appears in Slack
  • a row appears in a spreadsheet
  • a page is created in Notion
  • an email summary is generated
If the outcome is easy to verify, debugging is easier too.

​
Step 2: Create the workflow

Open Home and describe your use case in a prompt. Then send the prompt to create the workflow. Workflow Machine will use AI to turn that prompt into a starting workflow draft with the trigger and steps it believes fit your use case.

​
Step 3: Review and refine

After the workflow is created, review the draft in the editor and make any changes you want. You can adjust the trigger, reorder steps, update mappings, remove anything unnecessary, or add extra steps yourself. The AI-generated version is only a starting point. Keep the first workflow simple. Focus on making one path work clearly from trigger to outcome before adding extra branches or logic.

​
Step 4: Run a test

Testing is where the workflow stops being theoretical. Run a test to:
  • confirm that the trigger is valid
  • see whether each step executes in sequence
  • inspect outputs at each stage
  • catch field or auth issues before publishing
If the test fails, isolate the earliest failing step and fix that before changing anything downstream.

​
Step 5: Publish the workflow

When the test run looks correct, publish the workflow. Publishing makes the current version live. After that, real trigger events can create runs. A good habit is to publish only after:
  • the trigger is confirmed
  • the critical fields are mapped
  • the connection is known to work
  • the outcome is easy to verify

​
Step 6: Verify a real run

After publishing, trigger the workflow with a real input and review the run history. Look for:
  • whether the workflow started as expected
  • whether each step completed
  • whether the final output matches the intended result
This final check is important because real inputs often surface issues that test inputs do not.

​
Keep improving from there

Once the first version works, you can add:
  • better branching
  • more robust data handling
  • additional integrations
  • AI steps for summarization or extraction
  • template reuse for similar workflows
Start simple, get one successful run, then iterate.