AI myths and essential truths: What it can (and can’t) do for the insights industry

Depending on who you listen to, AI is either about to replace insights professionals entirely or unlock a golden age of human-robot insights. As is typical with technological developments, the reality is less dramatic than many of the AI myths floating about, and far more useful. 

AI isn’t progressing in a smooth, predictable line. It’s advancing unevenly, excelling at some tasks while remaining fragile in others. Market researchers are encountering this, and it’s what many technologists describe as the ‘jagged frontier of AI’. 

Understanding this frontier isn’t about resisting AI or surrendering to it. It’s about using it well. Because when market researchers understand what AI can and can’t do, they can work with more confidence and deliver insights more efficiently.

Read more: Human Experience in the AI era: A guide for insights leaders

What the jagged frontier means

The jagged frontier describes a simple but important truth: AI performance varies dramatically depending on the task. 

The same system that can summarize hundreds of interviews in seconds may struggle with irony, cultural nuance, or emotionally complex feedback. It may accurately surface patterns one moment, but it may struggle with concluding an array of correlations. This is true outside of market research, too. ChatGPT can provide impressive Deep Research answers on pretty much any topic, but it can’t count. This inconsistency isn’t exactly a flaw, but a defining feature of today’s AI systems. 

For market research teams, this explains why AI can feel tremendously disruptive, and also not at the same time. It would be a mistake to assume that success with a tool in one area applies to the rest of the process, and that it implies the redundancy of a human to operate, moderate, and expertly interpret. 

Where AI performs well

AI brings real, tangible value to market research. 

It excels at processing large volumes of information quickly, summarizing lengthy transcripts and open-ended responses, and detecting recurring patterns across datasets that would overwhelm human teams. It handles repetitive tasks like scripting, routing checks, transcription, translation, and basic QA with speed and consistency that humans would struggle to replicate, reducing friction across global projects. 

Used appropriately, these capabilities remove long-standing bottlenecks. They accelerate delivery, reduce manual effort, and help insights teams keep pace with rising expectations and shrinking timelines. 

But speed alone does not guarantee insight quality. And this is where understanding the jagged frontier becomes critical. 

Where AI struggles

AI’s limitations matter most in the areas where market research creates its greatest value. 

It struggles with ambiguity, like questions that are exploratory, emotionally layered, or open to interpretation. It lacks true contextual understanding, making it difficult to assess why a finding matters within your stakeholders’ setting. It performs poorly when topics are genuinely novel or when historical data is limited, as it is restricted to its training data. And it cannot reliably make ethical judgments about sensitivity, appropriateness, or unintended consequences. 

This is why AI can be confidently wrong. It produces fluent output without understanding significance. It identifies signals without knowing which ones deserve attention. Without human judgment, AI-generated output risks looking authoritative while missing the point entirely. 

But those building AI for insights understand these limitations too. In market research, you’re generally not exporting your data sets to LLM chatbots. You’re looking at tools that automate the things AI is fantastic at, leaving more time for humans to do the emotional and contextually important analysis.  

Read more: Why AI built for research is built different 

Common myths that distort expectations 

Much of the anxiety surrounding AI in market research doesn’t come from the technology itself: It comes from misconceptions about what it can realistically do. When expectations are inflated, fear and disappointment tend to follow. 

Several myths appear again and again in conversations with insights teams:

AI Myths: AI will do everything 
AI is super intelligent, super capable and all-knowing. It will be able to take on an infinite number of tasks autonomously and undertake entire end-to-end market research projects without oversight. 

Reality: AI capabilities are very uneven. 
AI can exhibit very advanced performance in one area whilst being dangerously deficient in another. 

AI Myths: AI will replace insights experts 
All the constituent components of market research projects can be trained into AI models. Over time, fewer insights professionals will be needed until the point when they disappear altogether as models reach a critical performance threshold. 

Reality: Human validation is an essential safeguard 
Human oversight is still required, and it’s likely to remain so for an exceptionally long time. 

Errors, hallucinations, and non-compliance require human expertise, grounded in experience, to identify. This is true even for domains where AI has the most training data and the most adoption (such as software coding). 

AI Myths: AI will let stakeholders do all their own research 
Insights professionals won’t be needed because stakeholders in marketing, innovation, and other teams will simply ask AI to design and execute the research they want done. 

Reality: AI doesn’t eliminate the need for expertise 
Market research skills will remain critical even as AI undertakes a greater share of automatable tasks. 

People who understand how to ask appropriate questions, interpret data, and craft stories will be essential for turning said data into insights and action 

AI Myths: Synthetic data will substitute organic human feedback 
Synthetic data, digital twins and AI personas will be used instead of real humans to answer market research questions. 

Their speed and low cost will make them irresistible to impatient marketing teams and executives 

Reality: Synthetic data will have a growing role to play 
But the best results will come from properly deploying synthetic data in the most relevant areas, such as idea generation or initial validation. It will expand the number of questions that can be answered with data rather than function as a substitute for real primary research. 

And it will not provide answers for topics that are entirely new, creative, or exploratory. 

Discover more: Adapt to thrive: AI and the market researcher

How market researchers can work confidently with AI

Understanding the jagged frontier changes how insights professionals relate to AI. It’s something to understand as AI continues to evolve unevenly.  

Market researchers should be confident and competent with new AI technologies and adopt them appropriately to stay up to speed and competitive as this landscape evolves. Success comes from knowing when to trust AI with automation for efficiency gains and when to slow down and apply human judgment.

When approached this way, AI becomes less a source of uncertainty and more a collaborator. It handles speed and scale, while humans handle sense-making. The teams that thrive won’t be those who hand the whole workflow over to machines (or those who reject AI outright) but those who know exactly where the balance lies. 

Want to find out more? Visit our solution page to book a demo with one of our experts to see how these tools could help you and your market research.