Why AI Can't Spot What's Next in Culture

Over the past year I’ve coached a few marketing researchers and teams on using AI in their market research and operations. Often they indicated they liked using tools like Perplexity or ChatGPT for market research to get insights, but felt something was being missed. Yes, something is. And it’s quite big.

While AI tools like LLMs (LArge Language Models) such as ChatGPT, Claude, Gemini and Copilot are indeed helpful in research, they have a few significant limitations when it comes to market research. Specifically with regard to Cultural Intelligence (CQ).

AI tools are great for analysing cultural aspects that are already visible. That have happened. What they can’t do is spot culture as it’s forming. They can’t see cultural shifts, they can’t navigate the liminal spaces of culture as it forms.

AI tools like LLMs are statistical machines. Trained on what’s happened, with tons of academic and expert content. From the past. LLMs are rear-view mirrors. They can see documented culture. For marketers, that’s just too late. Especially for competitive advantage in a fast moving world.

Cultural intelligence (CQ) creates value before a cultural trend or practice has a name. It looks at the tension points where contradictory behaviours signal a transition that can leveraged for ad creative, campaign strategy and so on.

Then there’s “habitus”, which is our social rules for how we navigate the world. We know that the way we behave at a wedding is different from a funeral. It’s also the things people “don’t” say such as in focus groups or in surveys. Habitus requires being and operating in the physical world. AI can’t do this.

Cultural Intelligence, especially through cultural anthropology, requires the ability to sense tensions in societies and cultures. Culture emerges from binary oppositions working themselves out.

AI tools like ChatGPT, Gemini or Claude, as awesome as they are, are engineers. They recombine known elements. Culture requires “bricolage”, a term coined by anthropologist Claude Lévis-Strauss, which is how we take bits and pieces of our own and other cultures to assemble them into something new, emergent. As a cultural anthropologist I’m a “bricoleur”, someone who can sense which cultural materials are available for recombination *before* the recombination is complete.

So yes, LLMs are great for brainstorming. I use Claude and Gemini this way (ChatGPT sucks at this), and for analysing quantitative data and sometimes qualitative analysis through interrogation.

But through using them and through over a decade of providing cultural intelligence to clients, I also know where AI has its limits. CQ isn’t an AI thing. I’m not sure it ever will be.