Casper Henningsen says market research is a 40-year-old category built for an enterprise that no longer exists.
GetWhy's AI-moderated research on eBay yields a 15% GMV increase in targeted categories and multi-million euro cost savings over 18 months.
Henningsen says only about 3% of data stacks contain qualitative understanding, turning a research gap into commercial existential risk.
GetWhy bets human understanding is the agentic enterprise's scarce input
Market research is a 40-year-old category built for an enterprise that no longer exists, and GetWhy is betting its replacement is human understanding delivered at machine speed. AI moderators interview real people at scale, so brands get the why behind decisions within a day rather than in a quarterly report. As enterprises hand more work to agents, Henningsen argues verified human context becomes the scarce input.
Takeaways
Market research is built for a dead enterprise. Henningsen calls market research a 40-year-old category built for an enterprise that no longer exists. A decade ago a Fortune 1000 leader lasted about 14 years, he says, and now it is seven.
AI moderation blends depth with survey scale. GetWhy uses AI to orchestrate the research workflow, interviewing real people on video rather than in a room. Clients get hundreds of hours of coded, structured video instead of graphs of scores, so qualitative depth arrives at survey speed.
Enterprises choose between 10 percenter and 10xer. The first AI wave cuts cost across proposition shaping, demand space shaping, concept development and marketing. Banking the savings for the CFO makes a 10 percenter. Reinvesting cost-neutrally to innovate harder makes a 10xer, and Henningsen says the chance belongs to the 10xers.
GetWhy's eBay pilot shows the returns. The public eBay work, the first of a series, produced millions of dollars in research savings, a 15% increase in GMV in the targeted categories and a multi-million euro cost reduction over 18 months.
Data stacks lack human understanding. In the best examples Henningsen examined, only about 3% of the data stack held qualitative understanding and context. Faster decisions turn that gap into commercial existential risk, which is why he proposes a human layer under every commercial model.
Synthetic data needs verified humans behind it. Synthetic models can be the front end of enterprise agents, but Henningsen stays cautious. They must be built on grounded interviews and real truth data, with verified humans constantly updated at the back end, which Fabri compares to the composable models discussed on All In.
Understanding customers is the only edge. As building things gets easier, Henningsen says the only true source of differentiation is how well a company understands its target markets and customers and turns that into experiences that drive margins and revenue.
From prompting to autonomous loops
Today's models are mostly reactive, prompting and bringing something back. Henningsen says the true agentic modality is autonomous loops between models.
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
Coding agents are guided by consumer needs, while creative agents run loops until they meet a quality threshold.
GetWhy's pilots already run thousands of loops over hours, and human judgment steps in after the threshold is met.
Humans are placed where they are best, in judgment and strategic alignment, and enterprises not on that train will struggle to be competitive.