Insight doesn't need more beige. It needs more weirdos.

Insight is arguably having a beige moment. We all feel it. Open any deck, any report, any "provocative" finding, and you'll see the same five ideas in different fonts. How can this be, if the industry is full of smart, talented, highly experienced people?

Anthropology can help understand why this is happening to the industry, and what might be our way out of this sameness-creep.  One of the oldest observed biases in the discipline is a natural tendency for all human groups to establish social norms and, once rules are established, for behaviour to default to the mean and for the group to judge other groups by the standards of its own.

Anthropology calls this ethnocentrism. Originally applied to the basic moral idea that "my culture is better than yours", this is one of the reasons the discipline developed ethnography as method: to counterbalance ethnocentric tendencies by taking seriously the task of seeing life through other people's perspective.

This principle can be observed when similar tropes are used to discredit or dismiss different styles of thinking as less sharp, less rigorous, less reliable. Also, if you believe your group's standards are correct, you'll naturally prefer, trust, and recruit people who share them.

Sociology calls this homophily: the empirical tendency for people to form ties with others who are similar to them in education, status, values, and background. Once the group reaches critical mass, the belief that "our way is the right way" never gets tested or challenged, because the group has already filtered out the people who'd test it with a different perspective and way of doing things - a self-confirming feedback loop.

What this reveals is an overall tendency for people to respond to uncertainty by looking for cues of social cohesion. As relational beings, identification reads as certainty, and that is what social groups and organisations default to at times of instability.

Before we ask is this true, we ask, often without noticing: is this person like me? In doing that, we can easily mistake confidence for evidence, legibility for insight. Organisations filter out anyone asking "weird" questions. Data points and solutions that read as "fluffy" rather than "robust" get dismissed, if not backed by a methodology to make them defensible. But the weirdo in the room may be seeing blind spots invisible to the rest of us.

This is not simply an anthropological phenomenon. Economist Lu Hong and complex systems scientist Scott E. Page are best known for arguing that functionally diverse groups can outperform collections of the highest-ability individuals at solving difficult problems.  In their computational experiment using simulated agents, a team of randomly selected participants mathematically outperformed a team of experts. The reason Team Weirdo got the upper hand is that, while tackling the problem from different perspectives and heuristics, they were making different kinds of mistakes. Past a certain point, experts converge. They all have the same instincts for what good looks like. When they're wrong, they're wrong together.

Diversity helps by preventing correlated error. Experts can be wrong in assuming that everything visible to them is the whole system, and their confidence can be produced by the absence of contradicting data.

Ultimately, diversity is what enables innovation. It fosters the kind of thinking capable of spotting white space and producing surprising ideas with real cut-through. If everyone thinks the same, speaks the same, and reacts in the same way, they will hardly ask questions that would lead to different results.

The industry didn't run out of ideas. But it might be running out of weirdos. Of people coming from different socio-economic backgrounds. Of people coming in sideways from anthropology, philosophy, journalism, art. This might have to do with recruitment becoming more standardised and increasingly AI-enabled. Or fewer people studying arts and humanities subjects. But I also suspect it’s more about who gets to stay weird once entering the industry. There is an expectation to conform to a certain language, logic, and set of metrics that can be at odds with their very instincts that would make them valuable counterpoints to the current orthodoxy.

Procurement frameworks make it harder for agencies to propose work that doesn't fit into a supplier questionnaire. Timelines have compressed to the point where "when can you turn this around" comes before "is this the right question we should be asking”. Stakeholders are under pressure to move from brief to decision faster than ever, and protecting time for sitting in the messy middle feels like an indulgence.

When insights are being optimised only for how well they perform in a boardroom, how fast they travel within the organisation, and how well they play on a screen - quick, directive, frictionless, actionable - even fieldwork gets shaped by the same pressure: interviews scripted tightly enough to make sure we get a usable clip, rather than left open enough to go somewhere unexpected. We miss the opportunities that only ever show up at the edges: where the weird, the messy, and the unfamiliar haven't yet been smoothed into consensus.

There is always going to be friction when divergent ways of thinking must come together to solve problems. The real test is whether organisations can resist the pull to erase that friction.

And this is a pull that AI and data-driven systems are exceptionally good at satisfying, which may be part of why some clients are moving so fast to embed them into new ways of working.

If we let it, AI will only exacerbate these tendencies, pulling everything toward the mean only faster, with the added danger that its flattening impact in the quality of thinking is harder to trace – but instantly felt. Remove the human moderator, the human participant, the human brain, and you don't just remove cost. You remove the edges: the weird people, the weird ideas, the weird interpretations never show up, because they were systematically filtered out.

Instead, we need to defend an architecture that harnesses dissonance, using it to propel our thinking forward rather than treating it as a hindrance. If we want insights to earn its seat at the table long-term, we need to stop optimising for being right the same way, and get better at being right in different, more interesting ways.