At time of posting, I’ve just finished an (online) lecture for the University of Bedfordshire. It was for the institution’s Marketing Club as well as students of the “Strategic Digital Transformation and the Future of Marketing” schedule.
In the session, I wanted to move beyond a familiar question: Can AI produce insights?
I believe that a more useful question is: Where can AI help us reach better insights, and where do humans need to lead?
My answer is the Insight Cycle: five stages that take us from an initial question to action. Each stage calls for a different balance of artificial and human intelligence.
1. Questions: humans lead. Before gathering data, we need to ask what we are trying to understand. That takes curiosity, open-mindedness, preparation, simple and open questions, and a willingness to listen. AI can help sharpen a question, but we shouldn’t outsource the thinking that frames it.
2. Evidence: let AI do the heavy lifting. This is where AI excels. It can search, compare and synthesise large amounts of material far faster than we can. But we still need to check what it produces. A neat summary may smooth away an outlier—and that odd detail might be the clue that matters most.
3. Insight: humans lead again. I define being insightful as reaching “a profound and useful understanding of a person, market, situation, or issue.” Data helps answer what? Insight answers so what? and opens the door to now what?
That requires more than pattern recognition. It draws on context, lived experience, empathy and the ability to connect familiar things in a new way. Unilever’s “Dirt is Good” is a powerful example: it changed the laundry conversation from treating dirt as a parent’s failure to seeing it as evidence of children playing and exploring.
4. Judgment: pressure-test the idea. An insight that sounds compelling still needs scrutiny. Are we clear about the decision? Have we considered genuinely different options? Is the evidence fit for purpose? Who is accountable, and how will we learn from the outcome? One useful technique is a premortem: imagine the plan has failed, then ask what went wrong.
5. Influence: keep the human voice. If we want an insight to inspire action, we must understand the people we are speaking to. My advice is to write the core story yourself, then use AI to help refine, tailor and create variations. Finally, test it with real people in the intended audience.
In the Q&A, we talked about students submitting near-identical AI-generated answers to the same brief. My advice was to do your own first pass and never accept the first answer uncritically. Ask AI to challenge your thinking, not replace it.
AI raises the floor: it makes powerful analytical work more accessible. Humans can raise the ceiling. The goal isn’t to choose one over the other, but to know when each should lead.
—
Many thanks to Giles Robertson from the University for inviting me along. I look forward to our next encounter in a few weeks’ time, when instead we’ll be focused on the Six Golden Rules of Data Storytelling.
—
The first draft of this blog was written by a combination of Granola’s transcript of my lecture and ChatGPT. It was then edited by real humans. A nice bit of HIxAI partnership, I’m sure you’ll agree.