AI Can Answer Anything. Knowing What to Ask Is Still the Job.

For years I worked on behalf of a woman I never met. She started a new diet every Monday. The research told us 83% of women did. 

The data could show me what she bought, when she bought it and how her behavior moved. It could not show me Friday night, the warm roll and salted butter in front of her, the promise she made herself that she would start again Monday. 

Data is what she does. Insight is who she is. The difference between them is whether you understood her. 

That gap has not closed. It has only become faster to reach, which is a different thing. Ask a model a generic question and you will get a generic answer, now in seconds and backed by more data than ever. A fast answer to the wrong question still costs a year. 

The advantage sits somewhere narrower than the technology. It is the question you ask before anyone opens a tool, and whether you know the person on the other side of the answer. None of that is new. It is still where most of the value is lost. 

The Dynamic Duo has changed. The job has not. 

In 2018, Wes Chaar and I called the partnership between the marketer and the data scientist the Dynamic Duo. AI has changed the speed and the scale of that partnership. It has not changed the work, which is still to solve the right problem together. 

The marketer starts with the outcome. What are we trying to achieve, and whose behavior has to change for us to get there? The data, the models and the technology only matter if they answer that. 

Wes taught me years ago that AI is only as good as the data beneath it, and that neither one replaces judgment. 

What has changed is where the data lives. It is scattered now, across channels, platforms, systems and companies, so the duo has had to widen to take in Data and Technology. Marketing still defines the growth question. Data and Technology build the intelligence the business can trust. AI has collapsed the time between the two. 

The answer comes faster. The insight does not. 

What AI hands back is time. Research and analysis that used to take months can take days, sometimes hours. That is real, and I would not trade it. 

Insight still sits one layer down. Which signals matter? What tension sits underneath the behavior? What should the business do differently because of it? 

With Drop a Jean Size, the opportunity was never another product. It was what sat between what she wanted for herself and what she did on a Friday, and what a brand could honestly offer her there. A model can find the pattern and test the hypothesis. Someone still has to ask who is on the other side of it. 

And none of it is worth anything to the business until it changes what the business does. 

Where most good insight dies 

At Catalina I led marketing and communications through a post-bankruptcy turnaround. A company coming out of bankruptcy does not have a marketing problem. It has a belief problem. The clients who stayed want to know why they should keep staying, and the ones who left are not returning calls. 

I call the distance between what a company is worth and what the market believes it is worth the Narrative Gap. 

Closing it takes more than a good story. We said Catalina’s near-real-time data asset would be worth more, not less, as the business moved into digital and retail media. Nobody owed us that. It became true when the products shipped, the clients adopted them and retail digital sales grew 526%. Belief followed the evidence, in that order. 

At NielsenIQ the same problem arrived in a different costume. We were telling an AI-first growth story to the public markets, which meant showing that AI was changing our own business before we asked clients to trust it with theirs. 

The lesson from both is the same. Move from tell me to show me. Show me the solution. Show me the customer using it. Show me what it is worth. 

A narrative earns attention. Proof earns belief. 

Agentic commerce changes the question, not the discipline 

The next shift is already in the cart. In May, NIQ’s agentic commerce tracker put the share of US consumers who had used at least one AI tool to shop in the past month at 42%. By late September it was 51%, with product recommendations the most used at 20% and shopping assistants at 16%. A majority, inside two quarters. 

Which raises the question no tool has answered: What will consumers trust AI to decide for them, and what will they insist on deciding themselves? 

I will happily let an agent reorder something routine. Would I hand it what I feed my family, what I put on my skin, where I travel? Not in the same way. That line moves by category, by generation, by price, and by how much of her identity rides on the choice. 

So the growth question changes shape. Persuading her was the old job. The new one is becoming the brand she, and eventually her agent, trusts enough to choose. 

Better questions, better decisions 

The temptation, as AI settles into how companies run, is more content, more analysis, faster versions of what we already do. The opportunity is better decisions. 

Three questions are worth asking before the work starts. 

What outcome are we trying to change? Start with the growth opportunity, not the technology or the data you happen to have. 

What do we actually understand about the behavior driving it? Go past what people do, to why they do it and what would move them to do something else. 

What will we do differently, and how will we know it worked? Name the decision and the measure before anyone leaves the room. An insight nobody acts on is a slide. 

The models will keep getting bigger and the answers will keep arriving faster. 

Every model I have worked with can tell me she starts on Monday. Not one of them has ever mentioned the roll. 

That is still the job. 

Sources:  

NIQ, “42% of Consumers Now Use AI Tools to Shop, NIQ Data Shows,” May 5, 2026. NIQ, “Majority of U.S. Consumers Now Use AI to Shop, NIQ Finds,” September 24, 2026. Both from NIQ’s monthly agentic commerce tracker, approximately 500 US consumers per wave.