There’s a natural human tendency to seek information that supports our internal beliefs and instincts, but in strategic decision-making, confirmation bias becomes a liability that needs active management. And this is also where GenAI can lead us astray.
GenAI models are trained with reinforcement learning from human feedback, which makes the model’s behaviour agreeable and non-confrontational.
The result? Sycophancy, where GenAI can act like a digital yes-man. While the technical side of this phenomenon is well documented, it poses a substantial risk in high-stakes business decisions.
Market research is grounded in objectivity, serving as the critical friend that identifies and challenges a flawed hypothesis—before you make a multimillion-dollar mistake.
GenAI sycophancy does the opposite. It takes your existing assumptions, wraps them in technical language and presents them back as data-backed analysis. But FMCG brands and governments don’t commission professional research for reassurance; they want the unvarnished truth.
A case study in GenAI plausibility: The e-scooter trap
Singapore’s e‑scooter market offers a clear illustration of this concern. Strict regulations were introduced in 2018, resulting in a tiny market where e-scooters are rarely seen in public spaces—a reality that becomes clear with our local, on-the-ground expertise.
When we asked a leading GenAI platform what the current market looked like in Singapore and whether it was attractive for e-scooter manufacturers, the first response sounded polished and plausible. The output was confident about growth, a strong regulatory environment and discerning consumers, concluding that Singapore offered a stable opportunity. All sounded sensible but was completely wrong.
Knowing its answer was incorrect, we intervened with a single corrective prompt, adding one piece of real-world evidence—a link to Singapore’s legal guidance. The GenAI response changed entirely, highlighting operational barriers, strict rules prohibiting use on footpaths, compliance burdens and a market too small to be commercially viable. The new conclusion was blunt: Singapore was an unattractive market for expansion. In its own words, “the juice simply isn't worth the squeeze.”
This is the sycophancy trap at work. Without expert human intervention—ideally, an insight engineer who knows how to interrogate GenAI’s responses—the initial, flawed advice would have been accepted as trustworthy.
The filter that keeps GenAI grounded in evidence
None of this means you should avoid GenAI for market intelligence, but it’s a powerful tool that must be guided by a trained expert who knows how to challenge its outputs.
53% of professionals said GenAI will impact their business in the next five years
Source: Euromonitor International, Voice of the Industry Survey 2025 (n=159)
An insight engineer assumes the model will try to agree and actively works against this tendency. That might include using adversarial prompts such as “find all the data that contradicts my hypothesis” to surface contradiction or tension in the information.
This cross-examination is now an essential skill and creates a clear choice for business leaders:
- Build internal insight engineers who combine deep subject matter expertise with the strategic fluency to challenge GenAI outputs constructively
- Partner with market research organisations rooted in objective, independent analysis
With GenAI easily able to serve up a smooth and believable answer in seconds, genuine critical thinking has never been more valuable. Market research is the filter that ensures integrity, keeping decisions grounded in evidence.
How can you guard against GenAI sycophancy? Tune in to this Opportunity Minded episode for practical tips on using GenAI with intention for market research.
