I built an AI outreach system for a client's business community.
Most of our members tend to procrastinate in reaching out to their warm contacts. Others struggle to remain consistent. Some would forget to follow up.
So I decided to build an AI system to make the whole thing easier and more efficient.
And in my monthly AI workshop, I walked them through the whole process so they can install the system and run it every day.
The response has been incredibly positive. So I want to share with you the key things in the process.
I normally start my AI builds with a braindump. I told Claude what I wanted to build and how I envision it to work.
And after a few back & forth, we locked in the design, functionality and the overall plan for the build.
Btw, that's how I normally build. I brainstorm with Claude all throughout the process. We'd call each other out. It would surface any gaps or blindspots in my plan. And I'd give it feedback on whatever it recommends.
It's my thinking partner, not just my "developer."
People always say 'don't outsource your thinking to AI' which I 100% agree.
But I also believe that when used right, AI can actually sharpen our thinking.
Anyway...
There was a point in our back and forth when Claude sent this long paragraph, and it's filled with technical terms I don't understand.
Instead of nodding along and blindly moving to the next step, I asked. I stopped and asked for a plain English translation.
You see, if I approve something I don't understand, I won't know exactly how to maintain that system because I wouldn't know why something breaks.
Most advice about working with AI stops at 'ask good questions.'
But there's a follow-up move that most people tend to overlook: when the answer comes back and you don't fully get it, that's information you need to clarify.
So here's what to do:
Stop before approving. Don't move past the part you didn't understand.
Ask directly. "Explain that in plain English." Or my other go-to, "how would you explain that to a 10 year old."
Don't move to the next decision or step until you're clear.
If you don't fully understand how your AI system works, you're essentially hoping everything holds and AI did what it's supposed to do.
Ask. Test. And validate it works.
If you're at all curious about the AI Outreach System I built, I walked through the whole thing in this week's video.
Including the four assumptions Claude took apart before we built anything, the point when I almost bloated the whole thing, and the two steps most people skip.