Why human insight still matters in the age of qualitative research and AI
By Helen Jones, Head of Qualitative Research, Vision One

Artificial intelligence is changing qualitative market research, and there’s a lot to be excited about. From automated transcription and AI-assisted analysis to synthetic respondents, new technologies are creating opportunities to make research faster, more efficient and potentially more accessible.

But having spent more than 20 years working in consumer insight, I think there’s an important question we need to keep asking:

Is AI helping us understand people better, or simply process what they say faster?

Because there’s a big difference.

Qualitative research has never been just about collecting opinions or identifying common themes. At its best, it’s about uncovering the emotions, motivations, contradictions and often unspoken influences that shape human behaviour.

And that’s where I believe experienced researchers still play an essential role.

How is AI changing qualitative market research?

AI is already supporting different stages of the qualitative research process, from research design and recruitment administration through to transcription, analysis and reporting.

There are genuine benefits, particularly when we’re dealing with large volumes of interviews, multiple markets or extensive amounts of unstructured feedback.

AI can help us organise information, identify recurring topics, explore alternative hypotheses and find relevant evidence much more quickly.

But it’s important to distinguish what technology can help us do from where human expertise still adds value.

AI in qualitative research: technology versus human expertise

Research activity	How AI can help	Where human expertise matters
Research design	Generate potential questions, discussion guide structures and hypotheses	Understand the business problem and choose the right approach
Recruitment	Support screening and administration	Assess participant suitability, authenticity and sample quality
Interviews and focus groups	Transcribe conversations and organise responses	Build rapport, read the room, probe and manage group dynamics
Qualitative analysis	Identify themes, cluster responses and search transcripts	Interpret context, contradictions, emotions and unexpected findings
International research	Support translation and cross-market comparisons	Understand cultural nuance and local context
Reporting	Structure findings and support initial summaries	Challenge interpretations and translate evidence into commercial recommendations

The opportunity isn’t simply to replace time-consuming tasks with technology. It’s to free up researchers to spend more time on the thinking, questioning and interpretation that create real value.

What can human researchers uncover that AI might miss?

One thing I’ve learned over the years is that what people say is only part of the story.

Take creative testing, for example.

You present a new advertising idea to a group of consumers. They’ll usually find something positive to say about it. They might tell you they like the imagery, understand the message or think it’s a clever idea.

But there’s a world of difference between an idea that generates an immediate reaction — smiles, nodding, people leaning forward and wanting to talk about it — and one that’s met with silence or polite indifference.

On paper, the feedback might look surprisingly similar. But having been in the room, you know the response was completely different.

An AI tool analysing the transcript might identify similar positive themes in both groups. Even tools that analyse video may detect visible reactions without fully understanding their significance in the context of the conversation.

An experienced moderator can recognise that difference as it happens, explore what’s driving it and distinguish between polite approval and genuine enthusiasm.

Recognising the unexpected

Sometimes the most valuable insight comes from the one person who disagrees with everyone else.

In a focus group, it’s easy for a consensus to develop. But an experienced moderator will listen for the person who sees things differently and explore whether their perspective reveals something others haven’t considered.

Equally, a seemingly throwaway comment can open up an entirely new line of questioning, sometimes revealing an unmet need or opportunity that wasn’t part of the original research brief.

AI analysis is often useful for identifying recurring themes, but less frequent perspectives can be overlooked or given insufficient weight in an automated summary. And while AI can suggest follow-up questions, it isn’t necessarily equipped to recognise the significance of an unexpected comment in the moment, particularly when that significance depends on wider business or cultural context.

The skill isn’t simply noticing something unusual. It’s recognising why it might matter and knowing whether to pursue it.

Reading the room (and knowing when to stay quiet)

Good moderation is as much about listening and observing as it is about asking questions.

You notice when the atmosphere changes. When a previously animated conversation becomes subdued, or a particular topic suddenly generates a stronger emotional response.

And sometimes, the most useful thing you can do is say nothing.

Allowing a participant a few moments of silence can give them the space to reflect, reconsider or articulate something they haven’t really thought about before.

AI-led interviewing tools can generate probes and respond to participants’ answers. But deciding whether a silence signals uncertainty, discomfort, reflection, or simply a need for more time requires contextual judgement. Asking another question isn’t always the right response.

These small, often instinctive decisions can take a conversation somewhere much more revealing.

And that’s perhaps the most important distinction. AI can help us capture, organise and analyse what people say. Experienced researchers bring the judgement to recognise what matters, explore what lies beneath the surface and understand what it means for the business.

Where can AI add genuine value to qualitative research?

None of this means we should resist AI. Quite the opposite.

Used thoughtfully, AI has the potential to enhance both b2b and consumer research in several important ways.

• Faster data processing: AI can support transcription, categorisation and searching across large volumes of qualitative data, reducing time spent on repetitive tasks.
• Exploring themes and hypotheses: AI can help researchers identify potential patterns, compare responses and consider alternative explanations worth investigating.
• Supporting international research: Automated transcription and translation can improve efficiency across markets, although cultural context and language nuances still require careful interpretation.
• Making existing insight more accessible: AI-powered search can help organisations revisit previous research, locate relevant findings and make better use of their accumulated knowledge.

For me, the greatest opportunity is what we do with the time these tools can save.

If AI allows researchers to spend less time organising information and more time thinking critically about what it means, that’s a positive development.

But faster analysis doesn’t automatically mean better insight.

The risks of relying too heavily on AI analysis

One concern I have is that we risk losing some of the richness of qualitative research if we become too reliant on automated summaries.

When you spend time immersed in interviews and conversations, you begin to notice connections, tensions and contradictions. You develop a deeper understanding of the people you’re researching.

And sometimes the most important finding isn’t the one mentioned most frequently.

There are several risks that researchers and businesses need to consider:

• Losing context: Sarcasm, hesitation, cultural references and subtle shifts in meaning can be misinterpreted when responses are considered in isolation.
• Overlooking minority perspectives: Recurring themes are important, but an unusual response can reveal an emerging need or challenge an established assumption.
• Accepting plausible but inaccurate conclusions: AI-generated summaries and interpretations need to be checked against original research evidence.
• Introducing bias: Researchers need to consider how AI tools may reflect or amplify biases in the data or analytical process.
• Protecting confidentiality: Participant consent, data security, privacy and transparency remain essential when using AI-enabled research tools.

AI can help us navigate the evidence. But researchers still need to question it, challenge their own assumptions and decide what the findings really mean.

Can AI replace real qualitative research participants?

One of the more interesting, and potentially controversial, developments is the emergence of synthetic respondents.

These AI-generated participants are designed to simulate how particular customer groups might think, feel or respond.

The appeal is easy to understand. If businesses could explore customer reactions without the time and cost of recruiting real participants, the implications would be significant.

And there may be useful applications for early-stage idea exploration, hypothesis generation or testing research materials.

But there’s an important distinction between generating a plausible response and establishing what real consumers actually think.

Real people are unpredictable. They contradict themselves, respond emotionally and are influenced by experiences and circumstances that don’t always fit neatly into a model.

Synthetic respondents don’t provide direct evidence of those lived experiences.

For businesses making important decisions about products, brands or customer experiences, that distinction matters.

Synthetic research may serve as a supplementary tool, but its reliability must be demonstrated for the specific purpose. It shouldn’t automatically be treated as a substitute for research with real customers.

What should businesses look for when commissioning AI-enabled qualitative research?

For organisations commissioning research, the question shouldn’t simply be whether an agency uses AI.

It’s whether the approach will deliver reliable insight that helps the business make better decisions.

I’d encourage research buyers to consider five questions:

1. What decision are we trying to make? Start with the business challenge, not the technology.
2. Where will AI be used? Understand which parts of the process are automated and where experienced researchers are involved.
3. How are findings validated? Ask how AI-generated themes and interpretations are checked against original evidence.
4. Are we hearing from real customers? Be clear about the distinction between genuine participant research and simulated responses.
5. Who is interpreting the findings? Look for researchers who can challenge assumptions, understand context and translate evidence into commercially meaningful recommendations.

Ultimately, good research should do more than tell you what customers said. It should help you understand why it matters and what to do about it.

How Vision One combines AI with human expertise

At Vision One, we’re exploring how AI can support our qualitative research processes, improve efficiency and help our researchers focus more time on the areas where they add the greatest value.

Our approach remains grounded in understanding real people, asking better questions and delivering insight that informs business decisions.

We use a range of qualitative research methods from in-depth interviews, focus groups to ideation workshops, depending on the challenge our clients are looking to address.

Qualitative research service	How it helps businesses
Focus groups	Explore attitudes, emotional responses, group dynamics and reactions to brands, products and ideas
In-depth interviews	Understand individual experiences, motivations and complex decision-making
Online qualitative research	Engage participants across locations through digital interviews, discussions and diary studies
Ethnographic research	Observe behaviour in real-world settings and explore the gap between what people say and do
Co-creation workshops	Work collaboratively with customers and stakeholders to develop and refine ideas
Concept testing	Explore how new products, services, advertising and propositions resonate with target audiences

Technology can support many stages of this work, but it doesn’t remove the need for curiosity, judgement and experienced interpretation.

The future of qualitative research: better technology, deeper understanding

I don’t see the future of qualitative research as a choice between AI and human expertise.

The most exciting opportunity is to combine the strengths of both.

AI can help us work faster, handle more information and explore evidence in new ways.

But good qualitative research has always required more than identifying patterns or summarising responses. It requires us to listen carefully, question what we’re hearing, recognise what isn’t being said and sometimes follow the conversation somewhere we hadn’t expected.

And perhaps most importantly, it requires us to connect what we’ve learned about people to the decisions businesses need to make.

Because ultimately, the value of qualitative research isn’t measured by how quickly we produce the findings. It’s measured by how much those findings help us understand — and what we do with that understanding.

Hear about AI ethics in MRS in this post and video by Debra Harding

Find out more about Vision One’s qualitative research services and how we help organisations understand the people behind the data.

About Helen Jones

Helen Jones is Head of Qualitative Research at Vision One, where she leads qualitative research to uncover the human insights behind brand, customer and business challenges. Helen helps organisations understand people’s needs, motivations and behaviours, translating rich qualitative insight into clear strategic direction and actionable recommendations.