In this latest episode of the Data Malarkey Podcast, Master Data Storyteller, Sam Knowles, speaks to Louise Twycross-Lewis, managing partner and head of insight at media agency, PHD, about what insight really means, how AI is changing research, and why human experience still matters. For Louise, the distinction between data and insight is fundamental. Data explains what is happening. Insight explains what it means, and what someone should do as a result.
Brutal simplicity
That difference sounds simple, but getting from one to the other is often difficult. Researchers can easily become overwhelmed by the sheer volume of information available to them. Twycross-Lewis argues that strong research starts with clear objectives, well-defined hypotheses, and a firm understanding of the decisions the work needs to inform.
She recalls a Market Research Society training course where participants had to reduce lengthy research presentations to just ten slides. The discipline was uncomfortable, but valuable: identify the three things the audience really needs to know, then provide enough evidence to support them.
Human insight in the age of AI
That instinct for distillation has become even more important in the age of AI.
Twycross-Lewis already uses AI to analyse large volumes of open-ended responses and identify themes across datasets that would once have taken researchers many hours to process manually. Used well, these tools can increase both speed and coverage. But she draws a clear line around their role. AI should support human researchers, not replace their judgement.
This becomes particularly important when organisations experiment with synthetic data and AI-generated audiences. Twycross-Lewis argues that research still needs human experience to expose the quirky, unexpected truths that make an insight genuinely useful.
Quaker Oats
She cites Quaker Oats’ “Deliciously Ugly” campaign as an example. The idea emerged from a human observation that porridge is not especially attractive to look at. That throwaway comment became a creative insight. An AI system might have replicated it after the fact, but would it have surfaced it independently?
The same concern applies to bias. AI models trained on incomplete or unrepresentative data can amplify existing gaps, whether those involve gender, ethnicity, culture, or other forms of representation. Better AI therefore requires investment in better data, not simply faster analysis.
PHD’s “Culture Currents”
Twycross-Lewis also discusses PHD’s Culture Currents research, which explores shifts in British attitudes and behaviour. Recent studies have examined optimism and anticipation. The latter found that, despite the rise of instant access and “eventification”, people still place considerable value on looking forward to things. Indeed, anticipation itself can become part of the experience. That creates opportunities for brands to engage not only with events, but with the build-up to them.
Summing up
Across the conversation, one theme keeps returning: technology can make research faster, broader, and more efficient. But insight still depends on curiosity, context, judgement, and lived experience.
The smartest use of AI may therefore be to help researchers spend less time processing data (addressing “the what?”), and more time understanding what it means (moving onto the “So what?” that can fuel the “Now what?”).