AI in Market Research Australia | What It Can and Can’t Do

Can AI Replace Qualitative Research? What Insight Leaders Should Actually Do With It

No, AI can’t replace qualitative research, but it is changing where the effort goes. AI is very good at the mechanical parts of a project: transcribing interviews, sorting large piles of open-ended comments, and drafting a first pass at themes. It is much weaker at the part qualitative research exists for: working out why people do what they do, noticing what they leave unsaid, and judging which finding matters for a decision. The practical answer for insight leaders is to let AI speed up the groundwork and keep experienced people in charge of design, conversation and interpretation. Use AI to get to the draft faster, and people to decide what the draft means.

Can AI replace qualitative market research?

Not the core of it. Qualitative research is a conversation between a skilled person and another person, and its value comes from what happens in that room: the follow-up question nobody scripted, the pause before an answer, the contradiction between what someone says and what they do.

AI can summarise what was said. It can’t reliably notice what wasn’t, and it has no stake in whether your decision turns out right. That’s why the sensible question isn’t “AI or people?” but “which steps can safely be handed to AI, and which can’t?”

Importantly Qualitative research done by humans gets to the bottom of the messy stuff.  Asks questions that might not be apparent, understands the non verbal and gets to empathy in a way that AI can’t.

Where does AI genuinely help in a research project

Used well, AI earns its place in the steps that are slow, repetitive and low in judgement:

– Transcription and tidying of recordings, so analysts start from clean text rather than raw audio
– Sorting open-ended survey responses into rough groups, which saves days when there are thousands of comments
– A first-pass theme list that gives the research team something to challenge and refine
– Searching across projects, so a question like “what did we hear about price last year?” takes minutes instead of an afternoon
– Drafting early versions of summaries and reports for a human to rewrite

In each case, the AI output is a starting point that a researcher checks against the source material. Time saved here can be spent where it counts: on better questions and sharper thinking.

Where does AI fall short, and what are the risks of relying on it?

The risks are worth knowing before you buy or commission anything:

– Confident errors. AI can produce a tidy, plausible summary that quietly misstates what participants said. Without checking against the transcript, nobody would know.
– Flattening. AI tends toward the average view. The unusual comment that points to a genuinely new opportunity is exactly the one a summary is likely to smooth away.
– No context. It doesn’t know your category, your history with customers, or the politics inside your organisation, all of which shape what a finding means.
– “Synthetic respondents.” Some tools simulate research participants from existing data. They can help generate hypotheses, but they can’t tell you how real people will react to something new, because they’re built from what people have already said.
– Data handling. Recordings and transcripts often contain personal information. Where that data goes, and who can see it, needs a clear answer.

What should you ask a research partner like Ruby Cha Cha about how they use AI?

A good partner will answer these plainly and without jargon:

1. Which steps of the project use AI, and which don’t?
2. Who checks the AI’s output against the original material?
3. Where is participant data stored, and is it used to train anyone’s model?
4. How are participants told about it, and does the consent wording cover it?
5. Can I see how a finding traces back to what participants actually said?

At Ruby Cha Cha we are specific and we have written policies to back it up.  We also adhere to The Research Society Code of Ethics and the new AI guidelines.  If a research partnery is vague or uncertain, you’re probably dealing with a team that uses AI as a tool rather than a talking point.

Ruby Cha Cha is all about: AI drafts, people decide

We apply a simple rule in our own work: AI can draft, but only a person can give the verdict.

Every AI-assisted step has to pass what we call the verdict test. Would a wrong answer here change a recommendation? If not (transcription, tidying, sorting), AI is welcome. If yes (what a finding means, what the client should do, which of three options wins), a person who was in the room makes the call and can point to the evidence behind it.

It’s a deliberately boring rule, and it works because it puts the line where the risk is. It also keeps the conversation with clients honest: we can say exactly where AI sits in the process, and exactly where it doesn’t.

How does an AI-aware approach improve business decisions?

It gets you faster answers without making them shakier. When AI handles the groundwork and experienced researchers handle the thinking, you spend less time waiting for analysis and more time acting on it, with a clear trail from each recommendation back to the people it came from. That trail is what lets a leadership team trust a finding enough to commit budget to it.

FAQ

Is AI-generated research reliable?

• It’s reliable for mechanical tasks that a person can check, such as transcription and sorting comments. It isn’t reliable as a stand-alone source of insight, because it can state errors confidently and smooth away unusual views. Treat AI output as a draft that needs checking against the original material.

 How is AI used in market research today?

• Most commonly for transcription, sorting open-ended responses, first-pass theme discovery and drafting summaries. Experienced researchers still design the study, run the conversations and interpret what the findings mean for the decision.

 Are synthetic respondents a substitute for real participants?

• They’re modelled on existing data, so they reflect what people have already said, not how real people will react to something new. They can help you form hypotheses to test, but they shouldn’t replace talking to your actual customers.

 Does AI make market research cheaper?

• It can shorten the analysis stage, but recruitment, moderation and interpretation still need skilled people. Any saving depends on the project, and the better question is whether the quality of the thinking has been protected.