Resources · Free

Free resource · The framework

Not everything needs automation.

Most guides hand you a stack of tools and tell you to connect them. This one is about the decision that comes first: which part of your business you're actually building, and how far it needs to go. Sometimes the right answer is a written checklist, not an agent.

Start with the shape of the business

A tool can't tell you what your business is missing. That's the part you decide, and it comes before any question about which AI to use.

Every one-person business runs on four cores, whatever it sells. They're functions, not software — which is why they don't change when the tools do.

Audience

How people find you, and why they stay once they have.

Offer

What you sell, who it's for, and why it's worth the price.

Delivery

How the thing actually gets made and handed over.

Money

How it's priced, collected, and tracked.

Build where the business is stuck, not where the work is interesting. Those are rarely the same core, and the gap between them is where most solo businesses stall.

The five levels of intervention

Inside every core there are recurring jobs. For each one, exactly one of these five is the right answer — and the skill is knowing which.

1

Delete

The job doesn't need doing. Most lists of recurring work contain something that exists only because it always has.

Nothing downstream depends on it, and nobody notices when it's late.
2

Document

The job needs your judgment, so it can't be handed off — but it can stop being reinvented every time. Writing down the sequence is the whole intervention.

You do it well, but differently each time, and you can't say exactly how.
3

Systematize

The job holds the same shape every time. Fix the inputs, fix the output, and it stops depending on memory or mood. Still run by a person — you.

The steps are stable, and the result should look the same whoever runs it.
4

Delegate

The shape is stable enough to hand to AI, with a check on the way out. You stop doing the work and start reviewing it.

You can describe the job completely, and you'd still read the result before it goes anywhere.
5

Automate

It runs without you touching it. Reserved for jobs where the output can be verified and being wrong is cheap.

You've watched it work repeatedly, and a bad run costs you an apology, not a client.

Most people jump to five. That's the mistake the whole framework exists to prevent. The correct answer is usually two or three.

Every level you climb adds fragility — one more thing to maintain, monitor, and repair when it quietly breaks. Climb only when the level below has stopped paying for itself.

How to pick the level

Four questions decide it. Ask them about the job, not about the tool.

  • FrequencyHow often does it come back?
    If it's rare, don't climb. Building for work that happens twice a year costs more than the work does.
  • SamenessDoes it hold the same shape each time?
    If the shape moves, stop at document. You can't systematize a moving target, and trying is how systems rot.
  • JudgmentHow much of you does the output require?
    If it needs your taste, your read of a person, or your name on it, stop at two or three. Judgment doesn't delegate.
  • Blast radiusWhat breaks when it's wrong?
    If wrong is expensive, never reach five. Keep a person between the work and the consequence.

The ceiling

Some jobs stay manual permanently, and naming them is part of building the system — not an admission that it failed.

If a job runs on your taste, your relationships, or your accountability, it stays with you. The point of everything below it is to buy back the hours those jobs deserve. A business where the founder has been fully automated out of the work isn't leveraged. It's just absent.

The whole framework in one line: find the core that's stuck, list the jobs inside it, and take each one to the lowest level that solves it — no further.

Every week

One workflow.
Every Monday.

One AI workflow, one real use case, one thing you can run today. Free.

Start Your Upgrade →