Off Kilter 224: Beyond Average.
tl;dr: As AI commoditizes average, advantage requires thinking beyond.
This week’s edition began as a LinkedIn post. Afterward, I thought it was too important not to expand into a full Off Kilter. The picture above is a very average ChatGPT averaging of what an all-knowing AI looks like.
AI-induced Dunning-Kruger Syndrome.
Spend any time at all on LinkedIn, and it’s hard to avoid the conclusion that a growing majority of the posts represent what we might label “AI-induced Dunning-Kruger.” The mechanism is simple. The LLM hands you an authoritative-sounding answer within a domain you don’t know well, and the platform rewards you for posting it. The goal is marketing yourself; your confidence is machine-manufactured, and your reach is algorithmically dictated. Ultimately, this is less about a descent into charlatanism and more about people falling en masse into the trap of performative expertise without having the understanding necessary to recognize what’s happening.
In my own domain, I see a constant barrage of posts presenting the work of Professor Byron Sharp and the Ehrenberg-Bass Institute of Witches and Wizardry as some form of universal marketing “truth,” when, in reality, it’s a single theory among many. And, like all theories, it has blind spots and requires significant adaptation before becoming commercially useful rather than academically interesting.
Now, let’s be clear. Worshipping at the altar of Sharp isn’t a new phenomenon that arrived solely with the LLM. In fact, if there wasn’t already a large online corpus of information reinforcing the perceived credibility of Sharp and EB theory, the LLMs wouldn’t be presenting it to us as fact in the first place.
What is new is the speed, scale, and low cost at which AI systems are averaging out knowledge in exactly the fashion I described above, and the implications of this averaging run far deeper than just marketing theory.
Synthetic amplification.
Contrary to the confidence with which they present their answers, LLMs don’t, in fact, deal in truth; they deal in probabilities. Their answers regress to the mean of whatever they find most legible within the corpus of their training data. Put more simply, they seek out a coherent pattern of consensus, treat this average as “truth,” and ignore outlying signals unless explicitly told not to.
This consensus averaging then feeds on itself synthetically. The more the internet fills with LLM-generated posts positing any single theory as truth, the more the next generation of AI models learns to treat this as the default, creating a self-reinforcing loop that becomes very difficult to break.
Ultimately, such synthetic feedback loops form an epistemic monoculture, which is a fancy way of saying that a single, averaged perspective becomes inevitable when everyone sees and repeats the exact same thing.
This matters far beyond marketing, because the mechanism isn’t specific to marketing. It’s a property of how every LLM works. The same averaging is happening to anyone whose thinking starts as an AI first draft: the lawyer, the analyst, the doctor, the designer. The AI-generated posts on LinkedIn are simply the canary in the coal mine for the rapid collapse of every thinking category toward the mean. What’s interesting about this isn’t how AI raises the bar on average, which I think it does; it’s about who will choose to go beyond average to create advantage because, as every thinking category collapses around the mean, the only sustainable source of advantage will be the capacity to think beyond.
The AI of design.
Reverting to the mean is one way of looking at the trap Ferrari fell into with the Luce. I’d argue the biggest problem with the Luce has nothing to do with what it looks like, well, not directly. Instead, it was Ferrari’s decision to hire Jony Ive to design it.
Ive is a singularly successful designer. But his vanishingly rare superpower is setting the bar for average across an entire category. Not as a compromise, but as the most legible and resolved design experience, the one that sets an expectation others must at the very least match. It’s a skill worth literal billions if you’re operating at the volume scale of, say, Apple. However, the entire point of Ferrari is not to be average and not to operate at anything like Apple’s scale. Ferraris are exclusive, deliberately not for everyone, deliberately polarizing.
While Mark Ritson views the Luce problem as Ferrari walking away from its brand codes, that’s only partly correct. Ive didn’t abandon the brand codes of Ferrari, he translated them into an EV-specific form that, on average, will be maximally legible and thus attractive to as many people as possible. But, oops, doing so means it’s no longer a Ferrari; instead, it’s an EV setting the bar for others to match.
This illustrates the inherent challenge of using AI to dictate how we should think, rather than working the other way round. In this tale, Jony Ive is the AI of design. Sure, he’s setting the bar for average. But setting the bar for average is no good if you’re Ferrari, and all of your value rests on being something far beyond it.
Guiding the machine, rather than being guided by it.
This is why domain expertise and critical thinking are more important than ever in an AI world. Training yourself, learning continuously, cultivating curiosity by reading widely, building your expertise, and developing the capacity to interrogate new ideas while challenging received wisdom won’t lead to replacement by AI. Instead, these are the exact qualities necessary to enable something far more important: the capacity to guide and dictate the lens through which we want AI to “think.” Not to have it dictate our thinking, but for us to guide it in how it should think.
Here’s a direct example. After correcting Claude once too often for spouting EB theory as settled fact, I built an empirical marketing skill within it that wasn’t grounded in “How Brands Grow” but in a more complete yet far less-cited work, “Empirical Generalizations About Marketing Impact,” edited by Dominique Hanssens. This is the same LLM, but with very different guidance. I then asked it to critique Sharp’s theories through this new lens. Here’s what it had to say:
“Sharp is mostly right about the thing he studied and overconfident about how far it travels. The core findings he popularized aren’t really his and aren’t controversial: double jeopardy, the Dirichlet patterns, the duplication-of-purchase law, the leakiness of loyalty. These replicate across decades, categories, and geographies.
Where I part company is the move from ‘these patterns are robust’ to ‘these are laws.’ What Sharp actually documented is an equilibrium — the steady state of mature, repertoire, low-involvement categories where products have reached rough parity and competition has collapsed onto availability. That’s a huge and important slice of the economy, but it’s a slice. The patterns weaken or invert in subscription, services, luxury, B2B, and genuinely differentiated or disrupted categories. The universalism is the error, not the findings.
The deeper issue is that Sharp’s framework is descriptive, dressed as prescriptive. It tells you how brands grow within an equilibrium without asking whether the equilibrium is worth accepting. So you can follow it flawlessly and optimize yourself into a low-value corner: lots of cheap reach, mental availability maxed, and nothing that compounds into margin or pricing power. Useful as a floor, dangerous as a ceiling.”
Thinking beyond the average.
When everyone is fed and then accepts the same set of averaged answers, convergence becomes inevitable. That’s true whether you’re a marketer, a lawyer, a CEO, a strategist, or a designer. However, this convergence isn’t inevitable if we choose not to accept it.
What I want to leave you with has nothing to do with Ehrenberg-Bass theory or Byron Sharp, per se. Instead, it’s to observe that there’s a clear source of competitive advantage that emerges as LLMs collapse all thinking domains toward the mean.
A principle that’s easy to say and hard to do: In an AI-mediated world, advantages will accrue to those who are most willing to think beyond the averages their competition is settling for.