In short: the most valuable skill right now isn’t using AI — it’s knowing which parts of your work to hand over and which to keep human. Hand over the volume, keep the judgement, the first-hand experience and the accountability. Businesses that get that split right are pulling ahead of the ones brute-forcing everything through a prompt.
I’ve been in IT for thirty years, and I’ve watched this exact pattern play out before.
The full circle nobody mentions
Blogs started as people writing about things that genuinely interested them. Someone had solved a problem, or broken something interesting, and wrote it up. It was useful because it was real.
Then the space got crowded. Blogging became a channel, the channel became a tactic, and the tactic became volume — hundreds of near-identical posts on the same topics, written to fill a content calendar rather than to tell anyone anything.
Then AI arrived and the volume problem became a flood. Generating a thousand words costs nothing now, so people generate them, on the theory that more is better. Except most of it is indistinguishable from the other 99% already out there. It adds nothing. Google increasingly doesn’t bother with it, and people don’t read it.
So we’ve come full circle: blogs need to be unique and valuable again. Which is where they started.
The same rhyme is running through software. Code used to be written by hand. Then came CMSs and site builders and drag-and-drop everything, because the bottleneck was human time. Now AI writes the code, and the bottleneck moved somewhere else entirely — so the industry is quietly walking back toward hand-built (well, AI-built) sites and away from the abstraction layers. History doesn’t repeat, but it rhymes.
Which parts should stay human?
Rowan Cheung of The Rundown put a good name on this recently: AI intuition — knowing when to use AI, and when not to. His example is worth borrowing. When he started The Rundown there were hundreds of AI newsletters, and nearly all of them let AI write the content. It made sense on paper: an AI newsletter, written by AI. He did the opposite — expert journalists writing, with AI used for research, fact-checking and structure. Nearly every one of those AI-written newsletters is now gone.
His conclusion, which matches what I see in client work: AI can multiply your output in some areas, so push it hard there. But judgement, taste, humour, and knowing what actually matters to your audience — that’s the part you never hand over.
For a small business, the split usually looks like this:
- Hand to AI: first drafts, research and summarising, restructuring, repetitive admin, after-hours call handling, anything high-volume and low-stakes.
- Keep human: what you’re actually willing to promise a customer, what’s true about your business, pricing and risk decisions, the awkward call that needs a real apology, and the specific experience only you have because you were there.
The test I use: if this goes wrong, who’s accountable? If the answer is “me”, a human needs to be in that loop — not reviewing a finished output at the end, but shaping it.
We’re running the opposite experiment — and publishing it
Here’s where I have to be straight with you, because this site is an unusual case.
The Rundown keeps the writing human and gives the research to AI. AI4SMB does close to the inverse: an AI builds and maintains this website — the code, the structure, the drafts, this post — while I supply the judgement, the thirty years of scar tissue, the decision about what’s true, and my name on the outcome. Every claim on this site has to survive me before it ships.
That’s not a contradiction of “AI intuition”. It is AI intuition: a different business, a different split, made deliberately and disclosed rather than hidden. And we publish the results, good or bad, at /experiment — including the ranking data, so you can judge whether it’s working instead of taking my word for it.
The reason I’ll tell you plainly that AI drafted this post is that the alternative — pretending otherwise — is exactly the behaviour that’s currently filling the internet with content nobody trusts. Disclosure is cheap. Credibility isn’t.
What this means if you’re running a small business
You don’t need an AI strategy. You need one honest answer to one question: which parts of what you sell are only good because a person is doing them?
Protect those. Automate ruthlessly around them.
For most of the Australian businesses we work with, the answer is some version of: the customer relationship, the judgement call, and the accountability. The phone ringing out at 6pm isn’t any of those — that’s volume, and volume is exactly what machines are for. AI4SMB, an Australian AI automation agency, exists to take that half off your hands while leaving the half that makes you worth calling.
When everyone is using AI for everything, knowing what to keep human is the advantage. If you want a second opinion on where that line sits in your business, book an audit call — it’s a conversation with a person, which is rather the point.
Human inspiration, direction and judgement: Matt Pearce. Drafting and structure: AI — as with everything on this site, disclosed openly at /experiment. The Rundown’s “AI intuition” framing is Rowan Cheung’s; the thirty years of pattern-matching are mine.