I’m building AI agents to see if I can create income that doesn’t require me to be at a laptop all day.
This is week 1.
No revenue yet.
Just the system starting to take shape.
Monday. 5am.
Hour and a half before the kids wake up.
Most days, that window disappears before I use it.
I used to spend it doing keyword research — opening tools, typing seed phrases, staring at spreadsheets, trying to figure out what was actually worth writing about.
By the time I had a direction, the morning was gone.
Not anymore.
What changed
I built an agent that does the first step automatically.
It pulls keywords, expands them, scores them, and drops them back into a sheet.
Slack pings me when it’s done.
I didn’t have to be there for any of it.
The output isn’t perfect. About 10–15% is noise.
But 85% is usable.
And the ones that surprised me were genuinely surprising.
Instead of “fire pit ideas” — it generated
“fire pit ideas for small patios under $300.”
That’s not a keyword.
That’s a real person’s real problem.
What it actually looks like
Eleven nodes in n8n.
Nothing fancy. Just connected steps.
I tightened the Claude prompt after the first run.
Better immediately.
It’s not done. It probably never will be.
But it removed the hardest part: starting.
The receipts
Agent receipt: 1 agent live. First workflow run complete.
Content receipt: 0 articles published. The foundation just got laid.
Financial receipt: $0. Still week one.
Time receipt: Built this in hour and a half before my kids woke up on a Tuesday. That’s the whole window. That’s enough.
It’s not much yet.
But it’s the first thing that’s run without me
and moved something forward.
Next week
Those keywords become briefs.
That’s when it starts to feel real.
If this is something you’re thinking about too — trying to build something that works without you having to be perfect every day — you’ll probably like where this goes.
I’ll send the next one Friday.