My first AI plan was embarrassing — in the opposite direction from what you'd expect.
Not too small. Too big. 77 agents across 10 business ideas. An agent for everything: research, content, ads, sales, support, finance, hiring, ideas I hadn't even started yet.
It looked visionary on a whiteboard. It was unbuildable in a real week, with real customers and a real team.
So I threw it out and rebuilt with one rule.
An agent earns its place only if it makes revenue or saves 5+ hours a week.
That filter killed most of the list. What survived runs my businesses today.
What survived
- Lead qualification — every enquiry tagged and sorted before the team starts.
- Follow-up drafts — the 2nd and 3rd touch, where most sales actually happen.
- Review collection — happy customers asked at the right moment.
- Weekly reporting — leads, sources and spend on one page every Monday.
- Content drafts — written from my voice notes, approved in one sitting.
Five systems. All boring. All close to money.
What got parked
Whole business ideas, not just agents. Real estate, a SaaS studio, dropshipping, affiliate sites — parked until the core businesses hit a revenue line I've set for myself. Not killed. Parked. The list exists so I stop re-deciding it every week.
The scoring sheet (copy it)
Before you build any automation — or start any new idea — score it:
| Question | Score |
|---|---|
| Does it bring in revenue within 90 days? | 0 / 1 / 2 |
| Does it save 5+ hours a week? | 0 / 1 / 2 |
| Does it touch a customer or lead directly? | 0 / 1 |
| Can a person approve its output in under 2 minutes? | 0 / 1 |
| Can it be live in 2 weeks? | 0 / 1 |
5 or more: build it now. 3–4: next month. Under 3: park it, and write down why.
The rule that keeps agents safe
Agents research, prepare, qualify and draft. A person approves anything a customer sees. That one rule is why my systems work and most "AI automation" doesn't — nobody wakes up to an agent that told a client something wrong.
Not sure what to automate first? That's exactly what the first call is for. Bring your list — I'll score it with you and tell you the one system worth building this month. Book a 30-minute call.