Stop Asking Customers What They Think.
July 29, 2026
Why AI is forcing CMOs to rethink what deserves human research.
Welcome to Edition 19 of The Takumi AI Brief, my weekly take on the AI shifts marketers should actually care about.
This week isn't about a new model.
It's about a habit marketing has carried for decades.
01 · The Trend That Actually Matters
Whenever marketing feels uncertain, we do the same thing.
Commission a study.
Run a survey.
Recruit a focus group.
Wait for the deck.
Somewhere along the way, we convinced ourselves that more research meant better decisions.
I'm no longer sure that's true.
Not because research stopped being valuable.
Because we've been asking humans questions that never needed humans in the first place.
That's why synthetic audiences have caught my attention.
Kantar, Qualtrics, YouGov and GWI are all building AI-generated research panels trained on years of behavioural data.
In controlled studies, some already mirror traditional research for early-stage testing.
Whether they're 90% or 95% accurate isn't the point.
The real story is what happens to the economics of a decision.
If testing fifty headlines costs the same as testing five, you stop protecting ideas.
You start challenging them.
If you can pressure-test pricing overnight, research stops being a project.
It becomes part of the workflow.
02 · The AI Shift to Watch
This is where the industry is having the wrong debate.
People ask if synthetic audiences will replace customers.
They won't.
They'll replace unnecessary research.
I've sat through enough focus groups to know customers are brilliant at explaining yesterday.
They're far less reliable predicting tomorrow.
That's not a criticism.
It's human nature.
Behaviour has always been a better teacher than opinion.
The average focus group tells you what people think.
The best marketing changes what people think.
Those aren't the same job.
None of this means we stop listening to customers.
It means we get disciplined about when we ask them.
Does every campaign need a recruited panel before the first creative route?
Probably not.
Does every naming exercise deserve weeks of fieldwork?
Probably not.
Should AI decide if your positioning captures the emotional reality of Ramadan, or feels authentic in Riyadh rather than London?
Definitely not.
AI is good at recognising patterns.
People are still better at recognising meaning.
That's a distinction worth protecting.
03 · If I Were in the CMO Chair
I'd rebuild research around two speeds.
Call it the Two-Speed Research Model.
Speed one belongs to AI.
Use it to kill weak ideas early.
Test messaging.
Explore pricing.
Compare propositions.
Challenge assumptions before they get expensive.
Speed two belongs to people.
Bring real customers into the room when trust, emotion, culture and behaviour are on the line.
Those decisions deserve human conversation, not synthetic confidence.
Stop measuring research by reports produced. Start measuring it by bad decisions avoided.
Did we kill three weak ideas before production?
Did we save six weeks?
Did we stop ourselves spending millions behind the wrong campaign?
That's what research is for.
Reducing the cost of being wrong.
Not filling out the board pack.
The CMO Takeaway
The focus group isn't disappearing.
It's becoming premium.
The real shift isn't AI replacing customers.
It's marketing finally getting disciplined about when customers belong in the room.
The marketers who win won't ask customers more questions.
They'll ask much better ones.
Mohit Lodha, Founder, Takumi AI