AI and Market Research: Why Experience Still Matters

Last year I was asked by my publishers, Kogan Page, to write a fifth edition of my book Market Research in Practice. I was initially daunted by the task. I had almost fully retired from running a market research company some ten years previously, and I wondered whether I was still sufficiently close to the industry to write a new edition.

After some thought, I decided to take up the challenge. I had the time to research and think about what new material should go into the book. The problem was that a new edition would require approximately one-third new content. Publishers understandably like to keep the length of a new edition similar to the previous one, partly for production and cost reasons. That meant that for every new idea I wanted to introduce, I would have to find words to remove from the existing book. I knew that would not be easy.

Fortunately, I still had many friends and colleagues in the market research industry, and I was able to ask for their views on issues about which I was uncertain.

One of the biggest issues was AI.

I was familiar with ChatGPT and, like many people, had been using it in my day-to-day life as a kind of much more sophisticated Google search. But I was less sure about how, and to what extent, AI was actually being used in the day-to-day work of market research consultancies.

My own dilemma: should I use AI?

There was another issue I had to confront: should I admit to the editor and my readers that I was using AI to help write the new edition?

Initially, it felt rather like an admission of failure. If I asked ChatGPT or DeepSeek to reduce a 3,000-word chapter to 2,000 words, thereby giving me another 1,000 words in which to introduce new material, was I somehow cheating?

I experimented with it.

The results were remarkably good. AI reduced the word count efficiently and, in my view, improved some of the writing. Eventually, I had to bite the bullet and use it.

There was, however, a line I did not want to cross. I could not, in all conscience, allow AI to generate the new material for the book. The new ideas and arguments had to come from me.

That did not mean that AI could not help with editing. I often write using dictation into a word-processing programme, which inevitably produces a fair amount of tidying up afterwards. I saw no reason why I should not use AI for this purpose. Indeed, it proved to be an amazing help.

This experience changed my attitude towards AI. I began to see it less as a threat to writers and researchers and more as an extraordinarily capable assistant.

What does AI mean for market research?

As I wrote the new edition, however, I became increasingly conflicted about the effect AI would have on the market research industry.

On the one hand, I could see how immensely useful it could be for exactly the sorts of tasks I was using it for: editing, summarising, generating ideas and providing a first attempt at a piece of work.

But there were much bigger implications.

AI could remove a great deal of the grunt work from market research. For example, imagine the task of taking a questionnaire in Word and converting it into an online survey. Much of the coding involved in that process could potentially be automated. That could have consequences for the people whose jobs have traditionally involved this kind of work – including my own daughter, who worked in the coding department of a market research company.

There is another, potentially much bigger, challenge.

Suppose a client wants to understand the size, structure and prospects of a particular market. In the past, the client might have commissioned a market research company to undertake the study, waited several weeks for the work to be completed and paid thousands of pounds for the resulting report.

Today, that client can ask ChatGPT a similar question and receive an answer within seconds and at virtually no cost.

Why would they pay for research?

This is the question that initially made me wonder whether AI might represent a serious threat to the market research industry.

Interestingly, I recently read an article in the Economist suggesting that employment in the market research industry had grown by 6% over the previous year. If that figure is correct, it suggests that the immediate threat may not be as great as I had feared.

But there is an important distinction here.

The fact that AI can produce an answer does not mean that it can produce good research.

AI as a very bright graduate

I became particularly conscious of this when I reached the chapter in my book dealing with questionnaire design.

As an experiment, I asked an AI programme to design a questionnaire examining people's breakfast behaviour on different days of the week.

Within seconds it produced a perfectly respectable first draft containing around 20 highly relevant questions.

I was impressed.

But I could also see the danger.

It would be very easy for an inexperienced researcher to accept the questionnaire as finished. In reality, it was a very good starting point from which a more experienced researcher could develop more insightful questions.

This gave me a useful way of thinking about AI.

AI seems to me to be rather like a very bright young graduate.

You can give AI a task and it will produce something remarkably quickly. Sometimes it will be surprisingly good. But a seasoned practitioner still needs to examine the work, challenge it, identify what is missing and decide whether it is actually fit for purpose.

That distinction is particularly important in market research.

AI cannot yet think the unthinkable

Where I believe AI still has a major weakness is in what I would call thinking the unthinkable.

Good research is not simply about collecting information. It is about interpreting that information and seeing possibilities that may not be obvious from the evidence.

That requires judgement, imagination and, often, years of experience.

Consider the position that Guinness found itself in 30 or 40 years ago.

A conventional analysis of the brand at that time would have identified its traditional customer base, it would have examined consumption patterns and demographic or geographic gaps where more Guinness could potentially be sold.

All of that is useful.

But what if the answer is not to sell more Guinness to the people who already drink it?

What if the answer is to make Guinness attractive to people who have never considered drinking it?

That is much harder to arrive at through conventional analysis.

Over the years Guinness successfully moved beyond its traditional image as a dark, winter drink associated particularly with older male drinkers. It developed a much more contemporary cultural identity, broadened the occasions on which the brand was consumed and reached younger consumers.

The important point is not that AI could never have suggested any of these ideas. It is that genuinely unconventional strategic thinking often starts with a question that does not naturally emerge from the existing data.

It requires somebody to say:

“What if we looked at this completely differently?”

That is where human judgement becomes particularly valuable.

The future of market research

I am certainly not suggesting that AI will have little effect on market research. Quite the opposite.

It will change the industry profoundly.

It will automate routine tasks. It will make research faster. It will make some forms of analysis much cheaper. It will help researchers design questionnaires, analyse open-ended responses, summarise interviews, identify patterns and produce first drafts of reports.

Some jobs will undoubtedly change, and some tasks that once required people will increasingly be automated.

But I do not believe this means that the market researcher becomes redundant.

If anything, it changes what we should expect from a good researcher.

The value of a researcher will increasingly lie not in producing information, but in knowing what questions to ask, deciding which information matters, challenging the obvious answer and turning evidence into insight.

AI can give you 20 good questionnaire questions.

The experienced researcher asks why those 20 questions should be there in the first place.

AI can analyse a market.

The experienced researcher asks whether the assumptions behind the analysis are right.

AI can identify what customers currently do.

The experienced researcher asks what they might do next.

And perhaps most importantly, AI can give you a very plausible answer.

The researcher has to decide whether it is the right answer.

That is why, for the time being at least, I think human judgement will remain one of the most important assets in market research.