Human intelligence is the context AI still cannot supply. Segment by values, not by age. And verify every insight back to a real person.

Mander started as a research fellow at Oxford, studying classical archaeology and completing a PhD in the social sciences on how families across the Roman Empire behaved. When the skills proved transferable, he went from predicting what Romans did 2,000 years ago to understanding how consumers now think, act and feel.

GWI has been around since 2009, founded to revolutionise market research by going digital with global reach. It interviews representative samples in more than 50 countries, more than a million individuals a year, asking about their lives, apps, brands and purchase intentions. With more than 15 years of data the trendability is the heart of what it takes to market, and that human data now feeds the AI tools the industry uses to decide.

Technically there is a cohort aged 17 to 29, but coherence inside it is close to zero. At one end, single people living with parents in full-time education; at the other, 29-year-olds married with families and careers. What connects them beyond the label is not much. The spotlight on Gen Z outranks any interest any previous generation ever got, driven by their digital native upbringing and being the youngest during events like COVID.

Differences come from life stage, technology adoption, mindset and values more than birth decade. Values, beliefs, mindset and interests define people far more than demographics like age and gender, which are usually marketing's building blocks. Becoming a parent at 20 or 40 unifies you with other parents more than any age marker. Mander would segment by interest, life stage and mindset first, then overlay age or generation to decide where and how to reach people.

The number of platforms and devices the young spread across is eye-watering, and total time online keeps going up, much of it concurrent rather than sequential. Five years ago the conversation was just social; now it is social versus short video versus longer video and everything else. The difficulty of reaching an audience across the plethora of places they hang out is more pronounced than it has ever been.

Twelve years ago GWI had no proprietary platform, delivering static data through a vendor plus PowerPoint reports. Now most people access via agentic interfaces, GWI's own agent or connections with ChatGPT, Claude, Gemini and Copilot. You used to go to the data and extract insight yourself; now the agent does the heavy lifting and you apply human context.

Mander asked the agent to find every place consumers contradict themselves, rank those contradictions by commercial importance and suggest follow-ups, even naming them. As a human analyst it would have taken days or weeks with no certainty of finding everything. The agent did it in minutes. But the output is only ever as good as the data in and the context given, so the agent has to know the personas and the industry it is serving.

A year ago the mood was fear and uncertainty; now it is outright skepticism about whether the data is trustworthy. Garbage in, garbage out. Synthetic respondents can be excellent or very limiting and biased, depending on context and source data. Use them for a quick steer, early process or a shortlist; be far more cautious for big budget or strategy decisions.

Mander frames three intelligence types coming together: artificial intelligence tools, which broadly do the same thing; proprietary intelligence, a company's own internal data; and GWI as the third pillar, human intelligence that gives context to the rest. The best predictor of how someone will answer, he adds, is how they have previously answered related questions, which is why 15 years of data is a goldmine for simulation.

There has been an explosion of capabilities, almost a land grab, and clients are exasperated at 95 per cent confidence claims while the black box stays shut. Mander gave an agent someone's digital life and asked their favourite flower from a list of five; it named one confidently, having run it repeatedly, with nothing in the source data to support it.

Challenge a tool and it backs down almost instantly with good point, or sorry. Those who do not push get the initial confidence, because most models were trained to affirm the views of the user feeding them. Mander says he regularly argues with AI tools that are not delivering what he wants. The demand now is transparency: scrutinise the vendor, the capability and the claims.

AI has taken insight from a small number of trained experts to hundreds, thousands or tens of thousands of users who are not from a research background and just ask a question to get an answer. Heads of comms, marketing and sales directors now self-serve what they once requested from specialists, giving GWI a goldmine of signal for its own product development.

It is democratisation with verification. Any user can immediately verify an insight and trace it back to real humans who said it in a real country. That need will keep rising: Mander says 99.9 per cent of surveyed professionals use AI and the genie is out of the bottle, so certainty over whether a figure is correct and verifiable only gets more important.

Careful about naming clients, Mander singles out the sneaker brand On for an amazing adoption journey, disrupting legacy players and creating buzz and exclusivity around a mainstream product. It is always telling, he says, when your own friendship group starts becoming customers one by one. To stay sharp he lives on the conference circuit, from South by Southwest to Cannes Lions, watching which topics and concerns dominate.

Two pieces. Evergreen: always look at what the data is telling you before you form your opinion, and do not go hunting for the story you want it to tell. Of the moment: there is no substitute for human experience and context, so develop skills that make you essential alongside AI. Be confident, because AI is changing the game, and work out how you remain not just relevant but necessary and important within it.