A year ago, we had perfect jobs at LinkedIn – fascinating work, great pay, and the stability we needed with small children and big mortgages.

Then we saw something that made us walk away from all of it.

What we saw was the convergence of two revolutions: generative AI and marketing science. For the past year, we’ve been mostly silent about AI. Why? Because we were too busy using it, every minute of every day. We’ve been partnering with the most elite marketing organisations to discover where AI fails, where it excels and – most importantly – what jobs it should do.
Today, we’d like to share our ‘three laws of leverage’. These laws separate organisations using AI for strategic advantage from those who are just playing with the newest tactical toy.

Think of it as a field guide from marketing’s future.

The conductor’s code: amateurs vs experts
Imagine yourself sitting down at a Steinway grand piano. What kind of music will you play? Should we expect Bach’s Goldberg Variations, or Chopsticks? If you can only play Chopsticks, is that because the Steinway is broken, or because you’ve never practised the piano?
Switch the instrument in this analogy for AI and you’ll find one of the great misconceptions at the heart of the AI conversation: the idea that using AI requires no skill or practice.

In fact, the amount of leverage that AI creates is proportional to the skill of its player.

We’ve probably logged over 500 hours practising the AI piano, and we’ve gotten pretty good, if we do say so ourselves. But we are mere toddlers compared to our chief technical officer, Brian Watroba, who might as well be Mozart. We can play a single melody; Brian can conduct a symphony. Like a master conductor who knows when to bring in each section of an orchestra, Brian can orchestrate a wide variety of AI models to play music far beyond our reach.

Different models at different ‘temperatures’ excel at different tasks. Think of them as sections in your AI orchestra.
Conductors mark their score with dynamics – from pianissimo (very soft) to fortissimo (very loud). AI experts mark their code by controlling each model’s ‘temperature’. The temperature determines how creative or conservative the outputs will be. A low temperature produces careful, predictable responses. Ask AI to tell you a bedtime story, and the low-temperature model will say: “Once upon a time, there was a princess in a castle.” If you set the temperature higher, you’ll encourage unexpected leaps, like: “Once upon a time, Cat Stevens flew an avocado to Saturn.”
Different models at different temperatures excel at different tasks. Think of them as sections in your AI orchestra. Some are like the brass section, powerful at computational performance and logical reasoning. Others are like the strings, bringing nuance and artistry to writing and creative tasks. Imagine you’re analysing concepts for a new ad campaign or a new product. The brass section can work at a low temperature (0.2) to calculate the financial value of relevant category entry points. Then you can turn to your string section, at a higher temperature (0.6), to brainstorm unconventional ways to win in those specific buying situations.

Today, most marketers think ‘AI’ means ‘ChatGPT’. But ChatGPT is just one of many models, and its temperature is pre-set to standardise the outputs. That’s a major limitation. So remember, when someone says “AI can’t do X”, what they’re really saying is “I can’t get AI to do X”.

They’re confusing an unskilled player for a broken piano.

Three revolutions are converging but only good marketers will benefit
The Picasso prophecy: right answers vs right questions
In 1968, an interviewer asked Pablo Picasso what he thought about computers. “Computers are useless,” he scoffed. “They can only give you answers.”
With all due respect to Pablo, he did a pretty shitty job anticipating the computer revolution. But he did a fantastic job anticipating the AI revolution. Which brings us to our second law: the leverage AI creates is proportional to the combined skill of the marketer and the programmer.
Instead of debating what AI should do, most of us are fixated on what AI can do. But we’re missing an essential truth: AI is the dumbest it will ever be, and it’s already as smart as many PhDs.

The real challenge isn’t what AI can do – AI can increasingly do anything – it’s what AI should do. As answers become abundant, the competitive edge will belong to the marketers who know how to ask smarter questions than their rivals.
Take brand health, for example. We’ve found that synthetic audiences can measure brand awareness with remarkable accuracy – correlations routinely above 0.80. But when we share this data with clients, we hit a more fundamental problem: awareness isn’t actually that useful. AWS has near 100% awareness among tech decision-makers, but that doesn’t tell their marketing team anything about how to drive growth. What matters is mental availability: does AWS come to mind when a startup needs to scale quickly, when an enterprise faces security challenges, or when a business outgrows its current cloud infrastructure? Pure awareness tells you if people know your brand exists. Mental availability tells you if they’ll think of AWS when it matters.

Now, without AI, you could maybe field one mental availability study a year, for a single market and category. With AI, you can field 100 studies a year, across 20 different categories and 50 different markets. But first you need to know to ask for mental availability instead of awareness, which means you need to have studied the literature on marketing effectiveness.
We are constantly reminded of that scene in the Dark Knight, where the Joker compares himself to a dog chasing a car. “I wouldn’t know what to do with one if I caught it!” AI will make data much easier to catch. But data is only valuable when it drives decisions.

The synthetic strategy: hard jobs vs easy jobs
Would you like to have a robot mop your floors or visit the Amalfi Coast for you? Would you rather AI write your advertising copy or run your market segmentation?

Our last law: the gains from AI are proportional to the difficulty of the marketing task.

And no task is riper for AI-generated disruption than… strategy. That’s right, we’re talking about the art and science of segmentation, targeting and positioning (STP).
‘It’s not a slam dunk’: How will AI impact segmentation and targeting?
Traditional marketing strategy has always been a costly, time-consuming nightmare. Consultancies charge hundreds of thousands of dollars for months of painstaking work. You need massive customer samples, extensive surveys and complex analyses to identify market segments. Targeting workshops, competitive mapping, endless positioning debates – the process is so unpleasant that most marketers skip strategy altogether and hop right into tactics.
If you give AI your hardest jobs – like segmentation, targeting and positioning – the gains can be revolutionary.
After six months and £600,000, BCG will hand you a static strategy. And, if your sales teams reject the targeting priorities or your positioning becomes outdated, you’ll do what most marketers do: put the 600-slide deck in a drawer and never think about it again.
Now instead of a man-made strategy, imagine a lab-grown strategy, built by an advanced AI system. The lab-grown version won’t just be faster and cheaper to produce – the time and money saved are staggering, but amazingly, that’s not even the main benefit – the real revolution is turning strategy development from an annual event into an ongoing conversation.

With AI, you can rapidly test different strategic combinations, experimenting with various segment definitions, target prioritisations and positioning territories until you find the most profitable combination of choices. Then you can take these options to internal stakeholders, incorporate their feedback and generate new variations in real-time. And when market conditions shift, you can evolve your strategy to stay current, with the click of a button.
Today, marketers are myopically focused on using AI for easy tasks like writing social posts. But if you give AI your easiest jobs, the efficiency gains will be limited. If you give AI your hardest jobs – like segmentation, targeting and positioning – the gains can be revolutionary.

The real bubble is about to burst
We love a hot contrarian take more than anyone.
But “AI is over-hyped” might be the coldest take in the marketing universe. We’ve spent the last year developing lab-grown marketing strategies for top brands. And let us tell you with extreme confidence that AI isn’t over-hyped. If anything, it’s under-hyped.

The bubble isn’t in AI – it’s in AI denial.

Most marketers are treating AI like a copywriting assistant instead of a strategic mastermind.

That’s the real bubble. And it’s about to burst.

If you’re not getting leverage from AI, consider our key principles: Are you approaching it like a conductor, orchestrating different models for different tasks? Are you using AI to chase the wrong data or to make the right decisions? And most importantly, are you deploying AI against marketing’s hardest challenges – like strategy – rather than its easiest tasks?
The orchestra awaits. Are you ready to conduct?

Lire l’article complet sur : www.marketingweek.com