[TL;DR: AI is making most SME marketing faster but not smarter. More ads, more content, more campaigns - none of it matters if the strategy is wrong. The businesses winning with AI have built a learning loop where every campaign makes the next one better. Here's the difference.]

Why AI Makes Most Marketing Faster But Not Better

AI tools haven't made most businesses better at marketing. They've made them faster at doing the wrong things.

More ad copy. More blog posts. More campaign variations. Produced in a fraction of the time, with a fraction of the thinking. The volume is impressive. The results, often, are not.

This isn't a criticism of AI tools. It's a criticism of how most businesses use them. They hand AI the output problem when the real bottleneck was always the input problem: what should we be saying, to whom, and why would they care?

If your strategy is flawed, AI will execute that flawed strategy at scale and at speed. That's not leverage. That's acceleration toward the wrong destination.

The businesses that are genuinely winning with AI aren't just producing more. They've built something different: a system where each campaign, each experiment, each piece of content makes the next one smarter. Compound intelligence, not just compound output.

Understanding the difference requires three frameworks from three very different thinkers. Bear with us.

The Arithmocracy Problem: When Data-Driven Becomes Data-Blinded

Rory Sutherland, Vice Chairman of Ogilvy, has a word for the people running most marketing departments: arithmocrats. Decision-makers who believe that superior analytical ability qualifies them to make every decision through numbers alone, and who systematically undervalue anything that can't be put in a spreadsheet.

AI has turbocharged the arithmocracy. Now you can generate 50 ad variations, A/B test them at scale, and optimise toward whatever metric you've defined as success. The process is rigorous. The outputs are measurable. The results can still be mediocre, because the metric you're optimising toward might not be the thing that actually drives growth.

Sutherland's core argument, backed by decades of behavioural economics research, is that human behaviour runs on psychology, not logic. The value of a thing is rarely its functional attributes. It's what those attributes mean to the person buying it.

An AI trained to write Google Ads copy will write ads that sound logical. Clear benefits. Strong CTAs. Relevant keywords. It will miss what Sutherland calls the psycho-logic - the irrational but entirely predictable psychological levers that make people actually respond. The ad that converts isn't always the most accurate ad. It's often the one that makes the prospect feel understood, or safe, or a little bit excited.

Producing more ads faster with AI doesn't solve this. If anything, it amplifies the bias toward logical-but-ineffective copy, because that's what's easiest to generate and easiest to justify in a debrief.

This doesn't mean data is useless. It means data should inform, not replace, the psychological insight that makes marketing actually work.

Byron Sharp's Warning: More Campaigns Isn't More Growth

Byron Sharp, professor at the Ehrenberg-Bass Institute, spent forty years building one of the largest empirical databases in marketing history. His conclusion, from data across 130+ brands in 13+ product categories, challenges most of what small businesses believe about how marketing works.

Here's why this matters for AI-driven marketing: at any given moment, roughly 95% of your potential customers are not in the market for what you sell. Byron Sharp and Les Binet have called this the 95/5 rule. Only 5% of buyers are actively shopping right now.

Most AI marketing tools are optimised for that 5%. Conversion-focused copy. Retargeting campaigns. Search ads that catch people mid-decision. These capture demand that already exists. They do nothing for the 95% who will become buyers later, often much later, and who will choose whichever brand they happen to think of when the moment arrives.

If all your AI effort goes into producing more conversion-focused content faster, you're getting marginally more efficient at harvesting today's demand while systematically ignoring tomorrow's. You're optimising for this quarter at the expense of next year.

The businesses that grow sustainably do both. They run activation campaigns (AI-assisted, conversion-focused) to capture the 5% who are ready now. And they build brand presence (broader-reach, distinctive, memorable) that earns mental availability with the 95% who will eventually need what they sell.

AI can help with both. But you have to know which is which, and what the right balance is for your category. Most AI tools won't tell you. They'll just produce more of whatever you ask for.

The Compound Engineering Insight: AI Should Learn, Not Just Execute

Here's where the thinking gets practical. Kieran Klaassen, general manager of Cora at the technology publisher Every, described something he calls compound engineering - a development methodology where each unit of work makes the next unit easier and more effective. Every bug fixed, every decision made, every pattern identified gets fed back into the system. The knowledge compounds.

His team runs multiple software products with a two-person engineering team. The output is disproportionate because they've built a learning loop, not just a production loop.

The same principle applies to AI-driven marketing. Most businesses use AI in a production loop: write the ad, publish the ad, check the numbers, write the next ad. Nothing from that cycle improves the intelligence that goes into the next cycle.

A learning loop works differently:

• Plan - Before creating anything, research the problem. What do we know about this audience? What has worked before? What are competitors doing? What does the data actually suggest?
• Work - Produce the asset, campaign, or content
• Review - Analyse the output against what you expected. Where did the prediction fail?
• Compound - Feed the learning back. Update your understanding of the audience. Refine the brief. Improve the input for next time.

Klaassen discovered that a 20-minute planning session before writing code routinely saved hours of debugging. The equivalent is true in marketing: a structured brief before writing ads, built on real audience research and honest review of past performance, will outperform a higher volume of quickly-generated but poorly-briefed creative.

What the Three Frameworks Agree On

These three thinkers come from very different places. Sutherland is a behavioural economist arguing against over-reliance on logic. Sharp is a marketing scientist arguing for broad reach and memory structures. Klaassen is a software engineer arguing for planning before production.

They converge on one point: the limiting factor in performance is not output volume. It's the quality of thinking that precedes the output.

What Sutherland and Sharp Agree On (Despite Disagreeing About Everything Else)

Sutherland and Sharp appear to have different theories of marketing. Sutherland focuses on psychological reframing, irrational signals, and the power of perception. Sharp focuses on empirical data, penetration over loyalty, and consistent distinctive assets.

But look at the underlying principle: both argue that most of the value in marketing is in things that are invisible to standard measurement frameworks.

Sutherland: "Logic always gets you to exactly the same place as your competitors." The creative leaps that create real advantage are the ones that seem irrational until they work.

Sharp: Most of the marketing budget should be reaching people who aren't going to buy today. You cannot measure the value of those impressions in this quarter's conversion data.

If you build your AI marketing system to optimise only toward measurable short-term conversion signals, you're building a system both Sutherland and Sharp would tell you is optimising toward the wrong thing. You're automating mediocrity.

What This Means for Your Business

If you're an SME running Google Ads and Facebook Ads, here's the practical implication of all this.

The businesses that will look back in five years and say AI transformed their growth won't be the ones who produced the most content. They'll be the ones who built a system that got measurably smarter with each passing month.

FAQ

Is AI actually useful for SME marketing, or just for big businesses with big budgets?

AI tools are disproportionately valuable for small businesses, not large ones. A solo operator or small team can now do research, planning, and content production that previously required a team. The key is applying AI to the right stages of the process. Small businesses often lack the structured planning that larger businesses take for granted: detailed audience research, documented learnings from past campaigns, explicit creative briefs. AI can fill those gaps at very low cost. A Google Ads campaign written after an hour of structured AI-assisted audience research will consistently outperform one written in five minutes of hasty production. The question isn't whether you can afford AI tools. It's whether you're using them where they actually create leverage.

How do you balance short-term conversion campaigns with long-term brand building when you have a small budget?

Les Binet and Peter Field's research across hundreds of IPA award-winning campaigns suggests a roughly 60/40 split: 60% of budget toward long-term brand building and 40% toward short-term activation. For SMEs with limited budgets, even a 70/30 split toward activation is reasonable when immediate cash flow is critical. But something should be going toward brand. For most SMEs, the most cost-effective brand-building channel is organic content: SEO articles, social presence, Google Business Profile, and consistent messaging across all touchpoints. This doesn't require a separate budget so much as a deliberate decision to create content that reaches and resonates with people who aren't ready to buy yet.

What does "mental availability" actually mean in practice for a small local business?

Mental availability, in Byron Sharp's framework, is the probability that your brand comes to mind when a buying occasion arises. For a local business, this is brutally practical. When someone in your service area has the problem you solve and turns to Google (or asks a friend, or scrolls their phone), do they think of you? Mental availability is built through consistent presence, distinctive brand assets (name, colours, tone, imagery), and reaching people across multiple touchpoints over time. For a local SME, this means Google Ads for in-market buyers, SEO content for people researching, a strong Google Business Profile for social proof, and consistent social presence for passive awareness. None of these channels works in isolation. All of them together build the probability that when the moment arrives, you're the name that comes to mind.

What's the single most valuable thing AI can do for my marketing right now?

Help you understand your customers better than you currently do. Most SME marketing is based on assumptions about what customers care about, written by people who think like the business owner, not like the customer. AI can analyse your Google Reviews, your competitors' reviews, search query data, and industry research to surface the exact language, fears, and desires of your target audience. That insight, applied to your ad copy, your landing pages, and your email sequences, will move the needle more than any volume of AI-generated content. Start with understanding. The production will follow.

Further Reading

• How Brands Grow by Byron Sharp - The empirical case for penetration over loyalty and mental availability over engagement
• Alchemy by Rory Sutherland - Why logical marketing consistently underperforms psycho-logical marketing
• Compound Engineering: How Every Codes With Agents - The learning loop framework applied to AI-assisted production
• The Long and the Short of It by Les Binet and Peter Field - IPA research on the 60/40 balance between brand and activation
• Avinash Kaushik's Occam's Razor - Measurement frameworks for understanding what's actually working in digital marketing