I Asked AI to Help Me Build a Kit Automation

Here’s where it went wrong.

Let me set the scene.

It’s a Friday afternoon. I’m building out a Kit visual automation as a portfolio piece, something to show prospective clients what a purchase-triggered onboarding sequence actually looks like under the hood. Straightforward enough. I know Kit well. I’ve built automations like this more times than I can count.

So I thought, why not use AI to help me document the steps as I go? Efficient, right?

It was not efficient!

Here’s what AI told me to do first:

"Step 1: Click the plus button and select Form Subscribe as your trigger."

Except Form Subscribe isn’t even a thing in Kit. The trigger is called Joins a Form. Different words, same confident tone, completely wrong.

Now, if you don’t know Kit well, you might have spent a good ten minutes hunting through the trigger options for something called Form Subscribe before quietly concluding that you were somehow looking in the wrong place, or that Kit had changed, or that you were just bad at this. The AI didn’t hesitate. It didn’t say “I think it might be called this.” It just told you, with absolute authority, to select something that does not exist.

And even if we set the naming issue aside entirely, a Joins a Form trigger fires when someone fills in an opt-in form. A freebie download. A waitlist. A newsletter signup.

It has absolutely nothing to do with someone buying something through ThirveCart.

When a customer purchases through ThriveCart, ThriveCart talks to Kit via the native integration and applies a tag to that subscriber. That tag application is what should fire the automation. There’s no form anywhere near this scenario.

I caught it. I pushed back. The AI agreed, apologised, and we moved on.

Then it told me to add this as step three:

"Step 3: Add Tag: access granted."

Again, sounds logical. You’ve purchased, you should be tagged as having access. Except the trigger, the tag applied by ThriveCart on purchase, already does that job. Adding a second tag action directly after the trigger to confirm the thing the trigger just confirmed is like locking your front door and then also locking it again to make sure.

Redundant. Unnecessary. Completely Baffling.

I caught that one too.

Then came my personal favourite. At the end of the automation, after both the yes and no paths had completed, the AI told me to add:

"A Goal Step, tagged as onboarding complete."

A Goal step. In Kit.

Kit does not have a Goal step. That’s ActiveCampaign. What Kit actually does is let you add a tag at the end of a sequence to mark completion, which is functionally the same thing but is an entirely different step in an entirely different place in the builder.

Three instructions. Three corrections. All from me.

Here’s the thing I want you to sit with for a moment.

The AI sounded completely confident every single time. There was no hesitation, no “I think this might be right” or “you may want to double check this.” It just told me what to do in a calm, authoritative tone, the same tone it uses whether it’s correct or not.

And if I hadn’t known Kit well enough to catch those mistakes, I would have followed every single one of those instructions and built something that either didn’t work at all or worked in a way that made absolutely no sense structurally.

Worse, I might have assumed I’d done something wrong when it didn’t work. Because the instructions sounded so plausible.

That’s the bit that concerns me most when I think about my clients using AI to troubleshoot their own tech. Not that AI is useless, it isn’t. I use it constantly and it genuinely helps me work faster. It’s that AI is confidently, fluently, very convincingly wrong sometimes, and if you don’t already know enough to spot the mistake, you won’t.

You'll just think you're bad at tech.

And you’re probably not bad at tech. You’re just being given bad instructions by something that sounds like it knows what it’s doing.

This is exactly why a second pair of eyes on your systems, a human pair, matters. Not because AI is the enemy. It absolutely isn’t. Knowing which instructions to trust, which steps to question, and which “Goal steps” definitely do not exist in the platform you’re using requires actual platform knowledge.

That’s what I bring. Not a list of steps copied from a tutorial. An understanding of how these systems actually work, what they’re actually doing, and why.

So if you’ve ever followed an AI tutorial, hit a wall, and quietly assumed you were the problem?

You probably weren't.

Gem Roberts is a Systems Strategist helping course creators, coaches, and membership owners understand and fix the tech that runs their business. If your systems are held together with good intentions and crossed fingers, let’s talk.

Comparison of systems strategist and tech VA roles with workspace background.

Systems Strategist vs Tech VA

Do you ask yourself the question where is someone to fix this, or what needs to be done to fix this? Asking either of those questions can have a very different outcome.

Frustrated woman with head in hand at computer desk.

Why Your Tech Stack Keeps Breaking

Are you fixing the symptom or the cause? You’ve built your systems and your business and have clients coming in, but if things keep breaking, it’s taking you away from what you want to be doing when taking a few more mins could mean the difference between a fix or a band aid.