Stop asking me.

I was setting up Google Ads campaigns recently with the help of AI, and somehow I ended up doing most of the work.

Click here. Open this menu. Copy that value. Send a screenshot.

It knew exactly what needed to happen, but kept walking me through doing it myself. At some point I got annoyed and asked a question I should have asked much earlier:

What do you need to do this yourself?

Turns out, API access.

If you're not technical: an API is a door into a product that software can use directly, without the screens and buttons made for people. Google Ads has one. Everything I was clicking through by hand, the AI could do by sending requests to it.

If you are: a developer token and OAuth credentials for the account. From there it wrote the campaigns, keywords, ads, budgets and bids as one file and built them through the Google Ads API.

I gave it access, and it went off and built three campaigns. When it finished, everything was ready to go, but paused until I approved it.

That's how I wanted it to work in the first place.

I've started paying more attention to this because it's surprisingly easy to spend your afternoon helping AI do its job. You give it a task, it asks you a question, you answer, it asks another. Before long you're just following instructions from something that's supposed to be saving you time.

Sometimes there's a good reason for that. I don't want AI spending money, sending things to clients, or publishing design changes without my approval.

But I don't need to approve every step that gets us there.

There's a second part to this, too. When I keep having to make the same correction, I don't want to repeat myself. I want that correction to become part of how things work next time.

That's increasingly how I approach using AI: give it what it needs to finish the job, decide where I actually need to be involved, and make sure we're not having the same conversation twice.

If you want help setting that up for your team, see how AI Enablement works, or let's talk.