What this automation does
Every time a new record is created in your Airtable base, this automation reads the record's text fields, uses AI to understand the content, and fills in tag, category, priority, and other classification fields automatically.
This works for any Airtable use case: CRM contact records get tagged by industry and company size. Content calendars get categorized by topic and content type. Project tasks get assigned priority and department labels. Bug reports get classified by severity and component. The AI adapts to whatever tagging schema you define.
Tools you need
- Airtable: Your database — needs Pro plan or higher for automations and webhooks
- OpenAI API: Reads records and assigns tags based on your taxonomy ($0.005-0.02 per record)
- Make or Airtable Automations: Triggers on new records, sends to AI, writes tags back to Airtable
How to set it up
Step 1: Define your tagging taxonomy. List every tag, category, and classification field in your Airtable base. Write 2-3 example records for each tag so the AI knows what to look for.
Step 2: Create a Make scenario (or Airtable automation) triggered by new records. Extract the key text fields — name, description, notes, or any free-text content.
Step 3: Send the record text to OpenAI with a system prompt containing your tagging taxonomy and examples. Ask for output as JSON with specific field names matching your Airtable columns. Request a confidence score for each tag.
Step 4: Parse the AI response and update the Airtable record with the assigned tags. For tags with confidence below 80%, flag the record for human review by setting a 'Needs Review' checkbox. Run the first 50 records and adjust the taxonomy based on AI mistakes.
Cost breakdown
| Item | Cost | Notes |
|---|---|---|
| Make | $10/mo | Starter plan covers most record volumes |
| OpenAI API | $3-$10/mo | ~$0.01 per record at 300-1000 records/mo |
| Airtable | $0 (existing) | Uses your current Airtable plan |
| Setup time | 25-40 min | One-time, including taxonomy definition |
| Total monthly | $13-$20/mo | Saves 1+ hour/day of manual tagging |
Frequently asked questions
You can configure it either way. In strict mode, the AI only selects from your predefined tag list. In discovery mode, you allow it to suggest new tags when it encounters records that don't fit existing categories. Most teams start strict and switch to discovery mode once the base taxonomy is solid.
For image-heavy records, you can use GPT-4 Vision to analyze images and assign visual tags. For document attachments, extract the text first (using a PDF parsing step) and then classify. The tagging logic stays the same — just the input format changes.