In this latest episode of the Data Malarkey Podcast, Master Data Storyteller, Sam Knowles, spoke to Zehra Chatoo, founder of Code for Good Now, a business on a mission to minimise bias, amplify culture, and unlock human creativity in a world increasingly shaped by AI.
Chatoo has spent her career close to insight. She worked for 13 years at media agency Manning Gottlieb OMD, rising to Executive Director, Head of Strategy, followed by five years as Head of Strategy and Planning at Meta. Today, alongside Code for Good Now, she is a board director and trustee of UN Women UK.
Defining insight
Her definition of insight has remained remarkably consistent throughout that career. She uses the acronym FREA: fresh, relevant, enduring, and actionable. An insight must offer a different perspective, matter to the problem being solved, contain an enduring human truth, and prompt action. A trend, observation, or interesting data point is not enough.
Getting from data to genuine insight requires what Chatoo calls “disciplined curiosity”. She is a strong advocate of the Five Whys, repeatedly asking why to move beyond what the data says towards what might lie beneath it.
That distinction becomes even more important with generative AI.
LLMitations?
Large language models, Chatoo argues, are extraordinarily powerful predictive machines. They can identify patterns and make predictions at speed. But prediction is only part of insight. The rest requires human judgement: questioning, critical thinking, curiosity, and a willingness to challenge apparent certainty.
Real insight can also be uncomfortable. Rather than immediately resolving ambiguity, people sometimes need to sit with the messiness of a problem. Chatoo believes that capacity will become increasingly valuable as AI makes other forms of cognition abundant.
Counteracting biases
Her work also explores the biases embedded in AI. She points to evidence of a gender gap in AI adoption, with women reporting concerns around trust, ethics, and being judged for using the technology. Code for Good Now’s own research tested that last barrier by showing people identical CVs, both explicitly created using AI, changing only the candidate’s name. The female candidate was judged less favourably on measures including competence and trust.
For Chatoo, responsible AI adoption therefore cannot simply mean teaching people how to use new tools. Organisations also need to consider culture, accountability, representation, and the conditions in which people feel able to experiment.
Permission to prompt
This philosophy informs Code for Good Now’s “Permission to Prompt” training – available directly from Zehra and her team, and – since September 2026 – via the Institute for Practitioners in Advertising (LINK). Rather than encouraging people to hand their agency to AI, it teaches co-creation: working with models while continually testing, questioning, and challenging their output.
Look to the future now? It’s only just begun …
What skills will matter as AI develops? Chatoo is wary of confident predictions about particular jobs. Instead, she highlights critical thinking, judgement, courage, community, and the ability to tolerate discomfort. And perhaps AI can help people spend less time doing something Chatoo identifies as one of the great abuses of data: producing enormous presentations that parade the analysis without revealing the insight.
Her alternative is refreshingly simple. Find what matters, put the reveal upfront, and then explain how you got there. The perfect, howcatchem structure for a murder mystery – popularised by Columbo and cheer led by the rather more recent Poker Face – rather than the comparative tedium – particularly for a naïve audience – of a whodunnit.
In an age of increasingly capable machines, the premium may not be on having more data or producing more answers. It may be on asking better questions, exercising better judgement, and knowing when the machine’s prediction is only the beginning.
—
The first draft of this blog was written by ChatGPT, using a transcript of the episode and an ever-refined prompt. It was then edited by real humans.