AI and the Intelligence Practitioner

Artificial intelligence is not coming to market research. It is already here , and the question is no longer whether to use it, but how to use it well.
The risk is not that AI replaces human judgement. The risk is that organisations mistake AI outputs for intelligence.

What AI can genuinely do
The capabilities are real and should be taken seriously.

Large language models can analyse thousands of consumer reviews, forum posts and social media conversations in the time it would take a human analyst to read a fraction of them. They can identify recurring themes, surface emerging sentiment, and flag patterns that would be invisible at the scale of human attention.

Synthetic data generation , the creation of statistically valid artificial datasets derived from real sample data , makes it possible to augment thin samples, model minority consumer segments, and test hypotheses without the time and cost of additional fieldwork. This is not speculation. It is in active use in research programmes across multiple industries.

Predictive modelling, powered by machine learning, extends the analytical range of what intelligence can do , moving from describing what happened to anticipating what is likely to happen next. Natural language processing enables the systematic analysis of qualitative data at a scale that was previously impossible.

Each of these is a genuine expansion of what the intelligence function can deliver. The question is what they cannot do , and why that matters.

The limits of pattern recognition
AI systems are, at their core, pattern recognition engines. They are exceptionally good at identifying regularities in large datasets. They are not, by themselves, capable of understanding what those regularities mean in a specific strategic context.

A model that analyses consumer sentiment around a brand can tell you that negative mentions have increased by 23% in the past quarter, and that the most common associated terms relate to pricing and customer service. That is useful information. But it does not tell you whether this represents a structural shift in brand perception or a temporary response to a specific incident. It does not tell you whether the pattern is concentrated in a strategically important segment or distributed across the general population. It does not tell you what to do.

That interpretive work , connecting data patterns to strategic context, weighing evidence against alternatives, framing a recommendation that a decision-maker can act on , requires human judgement. Not instead of AI. Alongside it.

The organisations that will use AI most effectively in consumer intelligence are not those that adopt it most aggressively. They are those that understand precisely where it adds value and where it does not.

A different kind of expertise
The emergence of AI changes what expertise in consumer intelligence means , but it does not reduce the need for it.

The practitioner who can only run a traditional survey and produce a frequency table is working with a diminishing toolkit. The practitioner who understands how to design an AI-assisted research process, interpret its outputs critically, and translate them into decision-relevant intelligence is more valuable than ever.

This requires a different combination of skills. Statistical literacy remains essential , perhaps more so, because AI outputs need to be evaluated, not simply accepted. Methodological judgement is needed to determine which tools are appropriate for which questions. Strategic understanding is required to connect intelligence to decisions. And analytical rigour is necessary to resist the temptation to treat confident-sounding outputs as validated conclusions.

The role is not diminished by AI. It is clarified. The parts of the intelligence process that were always about mechanical data processing are now augmentable by machines. The parts that require interpretation, context and judgement remain irreducibly human.

The question of data quality
One further dimension deserves attention. The quality of AI output is entirely dependent on the quality of the data it processes and the rigour of the questions it is asked.

Garbage in, garbage out is not a new principle. But it acquires new urgency when the scale and speed of AI processing can amplify the effects of poor data or poorly framed questions dramatically. A flawed prompt fed into a large language model produces a fluent, confident-sounding answer that may be substantially wrong. A synthetic dataset generated from a biased sample will reproduce and extend that bias at scale.

Consumer Intelligence applies AI as an enabler , one that amplifies analytical depth when used well, and can systematically mislead when used carelessly. The difference between the two is not the technology. It is the quality of the human intelligence directing it.

That, ultimately, is what makes the practitioner's role not a casualty of AI, but its necessary complement.

These ideas are often discussed with executive teams, institutions and organisations facing complex consumer decisions.