[TL;DR: Most marketers think AI will simplify their job. Mollick's research shows it amplifies both good and bad inputs equally. Binet and Field's 60/40 rule reveals the work AI can't automate. Sharp's brand science explains why strategic thinking matters more than ever. AI raises the stakes, not lowers them.]

AI Is Amplifying Your Marketing Decisions, Not Replacing Them

The most dangerous assumption in marketing right now is that AI will handle the hard parts. It won't. It will handle the execution. And it will execute your strategy at a speed and scale where mistakes compound faster than any human could manage alone.

That's not a criticism of the tools. It's a fundamental feature of how they work. The businesses getting the most from AI have figured this out. They use AI to move faster and test more. But the hypotheses being tested, the creative directions being pursued, the audiences being prioritised -- those decisions are still human.

What Ethan Mollick's Research Actually Found

Ethan Mollick, a professor at Wharton and author of Co-Intelligence, ran and studied dozens of real-world experiments testing AI's impact on professional output. The findings are more nuanced than the headlines suggest.

In a widely cited BCG study, consultants using AI performed 40% better on complex analytical tasks. That sounds like unambiguous good news. But Mollick's analysis of the same data revealed something crucial: the performance gain was entirely concentrated within what he calls the "Jagged Frontier" -- tasks where AI genuinely excels (structured analysis, synthesis, drafting, optimisation).

For tasks outside that frontier -- ones requiring contextual judgment, relationship reading, or genuine creative originality -- AI-assisted workers actually performed worse than those working alone. The AI gave them confident, plausible-sounding answers that were subtly wrong. And they trusted it.

This is the thing most AI marketing content doesn't tell you. Mollick calls the ideal model a centaur: human strategy, AI execution. The centaur model only works when the human provides genuinely good strategic inputs. If your strategy is weak, AI doesn't compensate for it. It accelerates it.

There's an uncomfortable parallel in digital marketing. Automated bidding on Google Ads learns from your conversion data. Feed it campaigns with a clear strategy, strong creative, and a compelling offer -- it optimises well. Feed it campaigns that are directionless -- it optimises toward nothing in particular, efficiently. The intelligence is real. The judgment about what to optimise for is still yours.

The 60/40 Problem Nobody Talks About

The research shows that marketing budgets should be split roughly 60% brand building, 40% activation for optimal long-term profitability. Activation (Google Search Ads, retargeting, promotional emails) drives short-term sales. Brand building (emotional advertising, broad reach, distinctive asset development) compounds over time and creates the conditions where activation actually works.

Here's the problem with AI and that split:

But the 60%? That's where it gets complicated.

Brand building requires emotional resonance. It requires saying something that feels genuinely true to a specific audience. It requires creative distinctiveness -- not just competent copy, but work that lodges in memory over time. AI can write fluent, professional marketing content at volume. What it produces trends toward the average of what exists. And in marketing, average is invisible.

Binet and Field's data is clear: the businesses that abandon brand building in favour of pure activation see short-term gains followed by long-term erosion. They win the conversion today at the cost of the category position that makes future conversions cheaper. AI makes it easier than ever to go all-in on activation. That's exactly why understanding the 60% matters more now.

Why Sharp's Research Makes This Even Clearer

Byron Sharp's foundational research at the Ehrenberg-Bass Institute established what actually drives brand growth: mental availability. Being the brand that comes to mind when a buyer enters the category. Not being the most loved brand, or the most loyal brand -- simply the one that surfaces first in the relevant moment.

Mental availability isn't built through better targeting. It's built through consistent presence, broad reach, and distinctive brand assets that create recognisable memory structures over time. Sharp's data, compiled across 130+ brands in 13+ product categories, shows that brands grow primarily by acquiring new buyers -- not by extracting more from existing ones.

This has a direct implication for how AI fits into marketing.

AI excels at finding and converting people who are already in market. Google Search captures demand that already exists. Facebook retargeting re-engages people who've already shown interest. Smart segmentation finds the most likely converters. All of this operates within a narrow, high-intent slice of the potential market.

But Sharp's research shows the vast majority of category growth comes from people who currently don't think about a brand at all. Reaching them, and making an impression that sticks, requires brand-level thinking -- broad reach, emotional resonance, consistent distinctiveness.

No AI tool can invent your distinctive brand assets. A memorable visual identity, a tone of voice that's immediately recognisable, a positioning that cuts through -- these require human creative judgment. AI can execute them consistently once they exist. It can scale what's already distinctive. It can't originate the distinctiveness in the first place.

The combination of Mollick, Binet and Field, and Sharp points to the same conclusion from three different angles: AI raises the strategic ceiling. The businesses that win with it are the ones that start with clearer thinking, not the ones that delegate the thinking to the tool.

What You Actually Need to Change

The practical implication isn't "use less AI." It's use it at the right layer.

Think about what Azeem Azhar documented in Exponential View earlier this year: Spotify's top engineers stopped writing code. They direct AI systems that write the code, while the engineers focus on architectural decisions and specifications. The AI handles production. The humans handle judgment.

The same logic applies to marketing. Your AI should be writing the ads, building the sequences, testing the variations, generating the reports. But the brief that drives all of that -- who you're reaching, what you want them to feel, what makes your brand distinctively yours, how this campaign fits your longer-term brand position -- that's the work that can't be delegated.

Three questions worth asking before you put AI to work on any marketing task:

What This Means for Your Business

The most important thing you can do to improve your AI marketing results isn't a better prompt library. It's clearer strategic thinking upstream.

Know who you're actually trying to reach and why they would care. Know what makes your brand distinctive and memorable. Know where you're investing in brand building versus short-term activation and why. AI can then execute that thinking at a speed and scale that would have been impossible three years ago.

Without that thinking, you're moving faster in a direction that hasn't been chosen.

The businesses that will look back at this period and say AI transformed their marketing won't be the ones who used it most. They'll be the ones who used it best -- as a centaur, not a replacement. Human judgment at the front. AI execution at the back. The speed gains are real. So is the responsibility for the thinking that drives them.

FAQ

Does AI actually improve marketing performance for SMEs?

Yes, with an important qualifier. Mollick's research consistently shows AI improves performance on tasks within its core capability zone -- structured analysis, drafting, testing, optimisation, automation. In the right context, documented productivity gains of 30-40% are credible. The limit is strategic judgment. AI improves execution; it doesn't substitute for knowing what to execute. For SME marketing, this means AI delivers the most value when the underlying strategy is already clear: who you're reaching, what makes your offer compelling, what your brand stands for. Without that clarity, AI speeds up activity without improving outcomes.

If AI can write marketing copy, why can't it build brand distinctiveness?

AI generates content efficiently, and much of it is technically competent. The problem is that distinctiveness, by definition, means departing from what's average or expected. AI is trained on what has already been produced -- its outputs trend toward the average of its training data. Genuinely distinctive brand assets require departing from the expected in a way that feels right for a specific business, at a specific moment, for a specific audience. That departure requires human creative judgment. AI can execute and scale distinctiveness once it's been established by humans. It struggles to originate it. A good test: if the output could have come from any competitor in your category, AI has averaged you into invisibility.

Should SMEs invest in brand building or performance marketing?

Both, in roughly the proportions Binet and Field recommend: approximately 60% brand building to 40% activation over the long term. For most SMEs, the balance skews heavily toward activation because it's immediately measurable and AI tools make it easy to optimise. The risk is real: activation only converts the demand that already exists. Brand building creates demand over time by building mental availability -- being the brand buyers think of when the need arises. An SME relying entirely on Google Search is competing for buyers who are already ready to buy, often against bigger competitors with more budget. Brand building builds the conditions where those same buyers think of you first, before they even run the search.

What should I actually change about how we're using AI for marketing?

Start upstream. Before generating any AI content, write a clear brief that covers: who this is for specifically, what they care about, what you want them to feel, how this fits your broader brand, and what makes this distinctively yours. Then use AI to execute against that brief at scale. For campaigns, have humans set the creative direction and strategic priorities; use AI to generate variations, test hypotheses, and scale what works. And invest deliberately in the brand-building activities that AI can support but not originate -- your positioning, your distinctive assets, your emotional advertising. Those are the inputs that determine whether your AI-powered marketing compounds over time or just spins faster.

Further Reading

• The Long and the Short of It -- Les Binet and Peter Field's IPA research on the optimal brand vs activation investment split
• Co-Intelligence: Living and Working with AI -- Ethan Mollick's research-backed guide to the centaur model and the Jagged Frontier
• How Brands Grow -- Byron Sharp's evidence-based framework for mental availability and what actually drives growth
• Consultants and AI: The BCG Study -- The Harvard/BCG paper behind Mollick's Jagged Frontier findings
• Exponential View #561 -- Azeem Azhar on what AI becomes at scale, and the Spotify engineering model