The past won’t tell you what will win the future

Real expertise in brands, and knowing how to build ideas with real people, just might. Four difficult truths about backing brands and propositions that don’t exist yet.

Imagine Duolingo didn’t exist, and you’d been asked to design the perfect language-learning app. The research would point you one way: learning should feel safe, patient and encouraging, because anxious learners learn less. Ask an AI and it would tell you the same, and it’s what everyone else would build.

Duolingo built an owl that nags you. It counts your daily streak, reminds you when you’re about to break it, and if you ignore it long enough, tells you it will stop bothering you, since its reminders don’t seem to be working. It’s now the world’s most downloaded education app.

Could anyone have seen that coming? We think so, but not from the data you already hold, or from simply asking people what they want. Ask learners, or a synthetic panel, and they’ll tell you they want a patient teacher and no pressure. That’s true as far as it goes. But ask instead about the last time they tried and gave up, and a contradiction appears. People really do want to learn a language. What stops them isn’t fear of failure. It’s that nothing makes today the day: the familiar “I’d love to, I just never get round to it”. The very lack of pressure they say they want is what lets them drift away.

That contradiction is the insight, and it’s exactly what gets flattened when you average what people say, whether by asking a synthetic panel or letting AI summarise your interviews. Spot it, and the rest follows. Look beyond the category, at fitness coaches, running clubs and games, and you find people who are quietly grateful to be pushed. Then you need to understand the value exchange, because a nag only works if every few minutes of effort pays you back with a small, visible win. Psychologists have long known that a sense of progress is one of the strongest motivators there is, and that people try harder the closer they get to a goal. And finally, how you show people the idea matters as much as the idea itself. Describe “an app that nags you” and everyone will say it sounds annoying. Show them what a year of small wins would feel like, and the moment the nudge kept them going, and the reaction changes completely.

Most big business decisions look like that: you’re asked to back a brand, proposition or message that doesn’t exist yet. The data you hold describes the past, and a quick round of new research won’t hand you the answer on its own. What gets you there is deep knowledge, a method for producing strong ideas, and a way of bringing people with you.

Here are four truths that explain why.

Every dataset a business owns, and every AI model, is a record of what has already happened. That tells you where you are, not where to go. When Pew Research tested AI “synthetic respondents” against real people, they were out by around 12 points on average, and on almost half the questions an answer real people gave never appeared. Another 2026 study found them too positive and far less varied than real people. Ask one today how to build a learning app and it will tell you to copy Duolingo, which is rather the point: it can only recommend what has already won.

Fresh research only helps if you’ve asked the right question, put the right ideas in front of people and know how to read what they say. In 1990 the researcher John Colwell pointed out that because there’s no obvious yardstick for judging qualitative work, weak work is hard to spot, and “good” drifts into “whatever the client finds acceptable”. AI has made that everyone’s problem, turning any set of interviews into a fluent report in minutes. And it tends to steer everyone to the same place: an award-winning NeurIPS 2025 study found that different AI models give strikingly similar answers to the same open-ended questions. That’s a strange place to look for an idea nobody else has had.

What makes the difference is knowledge built up over many projects, living in people and methods rather than spreadsheets. It means knowing the commercial challenge behind a brief. It means knowing what makes brands, propositions and communications win, often by borrowing from other categories. It means knowing people, through behavioural science and time spent watching what they do rather than what they say. And it means knowing the craft of building ideas and showing them so that people respond as customers rather than critics. Daniel Kahneman and Gary Klein showed that intuition can only be trusted when it has been learned through years of feedback. No AI can download that.

The hardest part is making sure the people who have to act were part of reaching the answer, seeing customers respond for themselves and coming out with a conviction that survives the first budget round. Research doesn’t set strategic direction; managers do. And the bolder an idea, the harder it is to commit to, and the more senior the sign-off it needs. An owl that nags your customers is a hard sell to a board, which is why the people deciding need to be in the room, not sent the deck.

We use AI every day for desk research, drafting and stimulus. But the judgement about what will win stays with experienced people, working with real customers and the people who’ll back the decision.

The patient teacher was the obvious answer, and the data would have backed it. The nagging owl took research that listened properly, a willingness to look elsewhere and the nerve to act on it. That’s what a bold decision really is.

KSBR Brand Futures. Customer strategies. Shared conviction. Bold decisions.