Airtable Automations: Know the Limits Before You Ask AI

Today you can build Airtable automations with just a prompt. You can ask Omni, Airtable's built-in AI, or connect Claude or ChatGPT to Airtable through MCP. So you might think you don't need to know how automations work anymore. But sooner or later AI will take you in circles, because it doesn't know the limits of the tool. If you're struggling to build an Airtable automation, you're probably up against the edges of what it was designed to do.

Disclosure: the Airtable link above is a referral link.

The Three Buckets: Build It, Bend It, Hand It Off

Every automation you want falls into one of three buckets:

• Build it. Airtable automations handle this well. Green light, build the thing.
• Bend it. It's a little fringe. You need a creative workaround, but Airtable can still do it.
• Hand it off. It's too complex for native automations and belongs in a dedicated automation tool.

The real question is how you tell which bucket an automation belongs in. Let's go bucket by bucket, using a client onboarding system for a small creative agency.

Build It: An Onboarding Task Generator

We always start with the data. A Task Templates table lists every onboarding task and which packages it applies to. A Starter client gets 6 tasks, Growth gets 9 and Retainer gets 12. When a client signs, the automation creates the right tasks for their package.

You could ask AI to build this. I build it by hand, because you have to check AI's work anyway, and you can't judge an automation until you know how to build one yourself.

1. Trigger: When a record matches conditions on Clients. Package is not empty, Signed Date is on or before today, Onboarding Tasks is empty, and Account Lead is not empty. Specific conditions stop the automation from firing early or twice.
2. Find records: on Task Templates, where Packages has any of the Package from the triggering record. A Growth client returns exactly 9 templates.
3. Repeating group: for each template found, create a record in Onboarding Tasks. Task comes from the current item, Client comes from the triggering record, Status is To Do, and Template is the current item's record ID.

Tip: if the trigger finds no matching record, create a test record that meets every condition and walk it through each step as you build.

Why Your Airtable Automation Isn't Triggering

This catches almost everyone. An automation won't run on records that already match its conditions when you turn it on. A record has to go from not matching to matching. In the video, the test client met every condition and nothing happened. Removing the package and adding it back fired the automation and created all 9 tasks. If you add an automation to a live base, existing records that already qualify will be skipped.

Bend It: Date-Based Reminders

Now a harder request: if the "Schedule kickoff call" task is still To Do three days after its due date, remind the account lead. Airtable has no trigger that fires when a date arrives, and no delay step. So we solve it in the data layer:

1. Add a formula field on Onboarding Tasks: DATEADD({Due Date}, 3, 'days')
2. Trigger the automation when that date is on or before today and the task is still To Do.
3. Send the reminder email to the account lead.

Why is this a bend? Date and formula triggers don't recalculate instantly. They refresh every few minutes while you're in the base and up to every hour when it's closed. A 9:00 AM reminder might land at 9:05. For a reminder, that's fine. But you're pushing Airtable toward its limits.

Hand It Off: 5 Signals You've Outgrown Airtable Automations

1. You're waiting on something outside Airtable, like a Stripe payment clearing.
2. You need a tool Airtable doesn't integrate with natively. Slack, Gmail and Outlook are built in. QuickBooks isn't.
3. You need to handle failures, like checking a payment again in an hour. Native automations can't retry.
4. Timing must be exact, like 9:00 AM on the dot.
5. Volume is high. Reprocessing hundreds of records every hour burns through your plan's monthly automation runs.

One or two signals: you can probably bend it. Three or more: move the automation outside Airtable. Your data stays in Airtable. Only the automation moves.

Our team mostly uses Make and Zapier, and n8n is also a strong choice. Make is usually cheaper and more developer friendly. Zapier usually costs more but is much easier for beginners.

Why This Matters More With AI

You can now ask Claude or ChatGPT to build Airtable automations for you. But AI often won't push back or look for the tool's limits. I've seen people waste countless hours asking software to do something it was never designed to do. If you're going in circles with AI, it's very likely one of the limits above. Knowing the boundaries is what keeps you out of the rabbit hole.

Need Help With Your Build?

Not sure which bucket your automation is in? Book a free 30-minute strategy call at gapconsulting.io/strategy-call. You'll talk with a real person on our team and leave with clear next steps.

Keep on building!