The Third Position on AI: It Doesn't Replace You and It's Not Just a Tool

There are two conversations happening about AI right now, and they’re both wrong.

Camp one says AI replaces humans. Automate everything. The machine does the work, you collect the cheque. Camp two says AI is just a tool — a faster spreadsheet. You tell it what to do, it does it, nothing changes structurally.

I’ve spent the last year building a business with Claude Code every day, and I’ve arrived at a third position that neither camp is talking about.

The Compression Curve

When I started working with Claude Code, every interaction was an instruction. “Write this function.” “Create this layout.” “Draft this email.” The model was a typist with a better vocabulary.

Six months later, I’m saying things like “build the roofingcrew.co.uk website following the existing pattern.” One sentence replacing forty manual steps. Not because the model got smarter — it’s roughly the same. Because I’d built the infrastructure underneath that sentence. The template exists. The colour mappings exist. The checklist is proven.

That progression keeps going. Instructions became tasks. Tasks became intent. Intent became posture — a mode of operating rather than a sequence of steps.

The levels accelerate. Each one comes faster than the last because the infrastructure is compounding underneath you.

Why the Wall Exists

Everyone has access to the same AI models. The difference isn’t the technology. It’s the territory.

The default approach is to treat AI like a vending machine. Insert prompt, receive output. When the output is mediocre, blame the model or write a longer prompt. What rarely happens is building the world that makes a short prompt powerful.

“Build me a website” means nothing to an AI with no context. “Build roofingcrew.co.uk following the existing pattern” means everything — but only because “the existing pattern” is real. The template. The checklist. The colour guide. The deployment process. Hundreds of small decisions already made and recorded.

A five-word instruction doing the work of a page-long brief isn’t magic. It’s compound returns on every small decision you documented along the way.

The Flywheel Nobody Sees

In the early days, I assumed I was in a race — building infrastructure downward while climbing abstraction upward, hoping the foundation could keep pace with the ambition.

That’s not what happens.

When I say “build roofingcrew.co.uk following the pattern,” the system doesn’t just build the site. It identifies what the pattern actually is, finds the gaps, and strengthens the template for the next build. The high-level command doesn’t consume infrastructure. It generates infrastructure.

The checklist gets tighter. The colour mappings get validated. The deployment steps get refined. The build is the infrastructure.

And each cycle makes the next one faster. The first site I built with Claude Code took days. The fifth took hours. By the tenth, the site itself is almost a given — the system’s attention goes to where the site fits in the wider ecosystem. Marketplace listing. Payment links. Search console. Deployment. The site is table stakes. The system around it is where the value moves.

The Bug That Became a Market Segment

Let me give you a concrete example.

I use Claude Code to research local businesses — auditing websites, checking reviews, understanding where they’re strong and where they’re losing opportunities online. During a quality check, the system flagged some junk data: email addresses that had been scraped from CSS files on template websites. Font designers, not business owners.

A machine that runs itself would have cleaned the data and moved on. I looked at those addresses and asked a different question: “Are those from website templates that got abandoned? Businesses that tried to build a site and gave up?”

Eight out of thirteen were on Wix. The template was there, but no real contact details had ever been added. These were businesses with five-star Google reviews and strong reputations — thriving on word of mouth despite having a website that does nothing for them.

That single question — born from understanding how tradespeople actually behave, not how data pipelines work — uncovered a market segment nobody designed. Businesses with proven intent (they tried to build a site), proven pain (they gave up), and proven quality (five-star reviews from real customers). Perfect candidates for a professional website.

The system found the junk data. The human found the gold in it.

The Vending Machine vs The Kitchen

A business that runs itself is a vending machine. It optimises for what it already knows. That works — until the market shifts, until something unexpected lands on your desk and someone needs to see what it means.

What I’m building is different. It’s a kitchen where the mise en place is always done, the prep work is handled, the quality checks run automatically, and the chef walks in and cooks.

The system handles research, auditing, reporting, site builds, documentation. It flags its own data quality issues. It updates its own records when it learns something new.

What it can’t do is taste. It can’t decide that thirty clients is the right number and thirty-one is too many. It can’t look at a bug report and see a business opportunity. It can’t choose what kind of company to build.

That’s not a limitation. That’s the design.

The Gap That Can’t Be Closed

The gap between someone who’s built this way for a year and someone starting today isn’t twelve months of work. It’s twelve months of compound acceleration. Every checklist, every documented decision, every correction — they didn’t just solve the problem in front of me. They made the next problem easier to solve.

You can’t close that gap by working harder or spending more money. You can’t buy infrastructure. You have to build it, brick by brick, through actual use.

Here’s the uncomfortable truth for the “AI replaces everyone” crowd: the infrastructure requires taste. It requires someone who understands their trade deeply enough to know what to encode and what to keep in their own head. A plumber with twenty years in the trade knows things about customers that no dataset captures. That knowledge, encoded into the right infrastructure, becomes the competitive advantage that no AI capability can replicate.

What This Means For Your Business

AI isn’t coming for you. It’s coming for the parts of your job you probably don’t enjoy anyway — the repetitive tasks, the data entry, the routine communications, the things that eat your day without engaging your brain.

What’s left is the good stuff. The thinking. The relationships. The craft.

But you have to build towards it. Not by buying an AI subscription and asking it to write your emails. By systematically documenting how your business works, what your standards are, what good looks like in your trade. By building the world that makes AI useful, rather than waiting for AI to become useful on its own.

The chef who preps properly cooks better. That hasn’t changed in five hundred years. The technology changes. The principle doesn’t.

Want to see what this looks like for a business like yours? I’ll audit your site for free — and show you what the infrastructure could do.