You're Using AI Ten Times More Than Last Year. You're Using It at Exactly the Same Level.

Every marketing team in Australia uses AI now. That sentence would have been remarkable two years ago. Today it's barely worth saying.

What IS worth saying: HubSpot's 2026 State of Marketing report found that 85% of those teams use AI for exactly the same thing. Content creation. Writing ads, drafting emails, producing social posts. The same tools, the same tasks, the same mediocre results at higher volume.

Here's where it gets uncomfortable. 52% of those same marketers now believe AI has made content so easy to produce that it's become less effective overall. They scaled the thing that was already commoditising. They poured petrol on a fire that was burning the wrong fuel.

The issue was never whether your business should adopt AI. That question is settled. The issue is that almost everyone adopted it at the same altitude and never climbed.

The Horseless Carriage Trap

Pete Koomen, a Y Combinator partner, wrote one of the sharpest essays on technology adoption in 2025. He argued that most AI applications today are "horseless carriages". They swapped a horse for an engine without redesigning the vehicle. His example: Gmail's AI draft feature promises to help you "get through email faster." You're still reading and writing emails. You're just doing it at higher speed.

This pattern repeats every time a transformative technology arrives. Early plastic products were designed to look like wood. Early cars looked like carriages without horses. Early spreadsheet users replicated their paper ledgers cell by cell, completely missing that the real value wasn't a faster ledger but pivot tables, scenario modelling, and analysis that paper could never do.

AI marketing in 2026 is in the horseless carriage phase. The technology is revolutionary. The application is not.

More usage. Same altitude.

The Magnitudes Problem

In February 2026, Azeem Azhar, founder of Exponential View, consumed 97 million AI tokens in a single day and published a framework that should change how every business owner thinks about AI.

His argument is elegantly simple: each order of magnitude of AI usage doesn't just mean more. It means different.

Most Australian SMEs are stuck at the "tool" level. They adopted AI, increased their usage dramatically, and never moved up a single magnitude. Only 6% of marketers have AI fully embedded in their workflows, despite over 80% feeling pressure to adopt it, according to industry research.

Your team is using AI more than ever. They haven't changed what they use it for.

Why More Doesn't Mean Better

When DeepSeek launched its ultra-efficient AI model in early 2025, Microsoft CEO Satya Nadella tweeted three words: "Jevons paradox strikes again."

He was referencing a 160-year-old economic observation: when something becomes more efficient, people don't use less of it. They use more. William Stanley Jevons noticed in 1865 that more efficient coal engines didn't reduce coal consumption. They increased it. Cheaper, more efficient use made coal viable for more applications, so total consumption exploded.

AI content creation follows the exact same pattern. The promise was efficiency: produce content faster, save time, redirect effort to strategy. The reality? A FirMA analysis found that organisations using AI to accelerate existing processes see a productivity gain of just 1.1% for the overall workforce. The time saved gets immediately reinvested into producing more of the same thing.

The numbers tell the story. An estimated 48% of brand social content is now AI-generated. 56% of consumers report seeing "AI slop" daily. Average sustained audience focus has dropped to 43 seconds. Everyone produced more. Nobody stands out.

We've written before about why AI-generated content blurs into sameness. The data is only getting worse.

UC Berkeley researchers tracked this phenomenon in real time. Their eight-month ethnographic study inside a 200-person tech company found that AI-augmented employees didn't reclaim saved time for rest or strategic thinking. They absorbed more tasks. Productivity rose, but so did voluntary workload. The researchers called it "workload creep": progressive expansion of task scope, accelerated pace, and blurred work-life boundaries. Several of the highest-performing employees quit. Not because of AI, but because the expectation that AI made them capable of doing more meant more was exactly what was demanded.

The trap is clear. You gave your team AI. They used it to do more of what they were already doing. Nobody asked whether those tasks were worth doing at all. As we've explored before, cutting costs was never the real opportunity with AI.

The Eurostar Lesson

Rory Sutherland, Vice Chairman of Ogilvy UK, tells what might be the most important business story of the last twenty years.

The UK government spent six billion pounds building new tracks from London to Paris to shorten the Eurostar journey by 40 minutes. An engineering solution to a time problem.

Sutherland proposed an alternative. For 0.01% of that budget, you could put WiFi on the trains. The journey wouldn't be shorter, but it would be productive. For a fraction more, you could hire supermodels to serve free Chateau Petrus for the entire ride. People would ask for the trains to be slowed down.

His point isn't about trains. It's about the difference between solving a problem and redefining it.

"Making a train journey 20% more enjoyable may cost almost nothing, whereas making it 20% faster might cost hundreds of millions."

This is exactly what's happening with AI in marketing. Businesses are spending their AI budget building faster tracks. More content, more posts, more ads, faster. But the problem was never "we don't produce enough content." The problem was always "the right people don't think of us when they need what we sell." Speed doesn't fix that. A different approach does.

Sutherland calls these psychological moonshots: small changes in approach that produce outsized results because they address the actual problem instead of the assumed one. The Uber map doesn't reduce waiting time for a taxi. It makes waiting 90% less frustrating. The London Underground's biggest improvement in passenger satisfaction wasn't faster trains. It was dot-matrix countdown displays at platforms.

Your AI could be a countdown display. Instead, you're using it to make the trains run marginally faster.

Penetration Thinking for AI

Byron Sharp's research at the Ehrenberg-Bass Institute established one of marketing's most counterintuitive findings: brands grow through penetration (reaching new buyers), not loyalty (getting existing buyers to buy more).

The data is emphatic. Binet and Field's analysis of 880 IPA effectiveness award papers found that penetration-led campaigns won 21 Gold awards versus just 2 for loyalty-led campaigns. The pattern holds across industries and geographies. Growth means reaching people who don't currently think of you.

The parallel to AI adoption is direct.

Most businesses are running an AI loyalty strategy. They took their existing workflows and used AI to do them faster or at higher volume. More ads. More emails. More social posts. This is optimising what exists. It's the loyalty play.

The growth move is an AI penetration strategy. Using AI to reach new capabilities you couldn't access before. Things that weren't possible, not things that were slow.

This isn't the horseless carriage. This is the car.

What Moving Up a Level Actually Looks Like

For a marketing team at an SME spending $3,000 to $10,000 a month on advertising, "differently" doesn't mean a massive technology overhaul. It means asking a different question.

Three practical shifts:

What This Means for Your Business

The uncomfortable truth: if your AI strategy is "use ChatGPT to write things faster," you're in the same bucket as 85% of every other marketing team. You adopted the technology. You didn't adopt the thinking.

The data says that matters. Organisations that use AI for genuine transformation see 2-10x productivity gains. Those that bolt it onto existing processes see 1.1%.

The difference between those two outcomes isn't budget or technical sophistication. It's whether you used AI to do the same things faster or to do different things entirely.

Rory Sutherland put it best: the biggest scope for progress comes not from improvements in technology but from improvements in how we think about the problem.

You have the technology. The question is whether you'll use it to build a faster horse or design the car.

Further Reading

• Magnitudes of Intelligence - Azeem Azhar's framework for how each 10x in AI usage changes what it fundamentally is
• AI Doesn't Reduce Work. It Intensifies It. - UC Berkeley's eight-month study on workload creep in AI-augmented teams
• AI Horseless Carriages - Pete Koomen (YC) on why most AI apps replicate the old world instead of building a new one
• From Potential to Profit: Closing the AI Impact Gap - BCG on why only 5% of companies see real AI returns
• Why the AI World Is Obsessed With Jevons Paradox - NPR on why efficiency increases consumption, not reduces it