AI and market research technology: Researchers have been lucky over the last few decades, but has our luck run out?

Part one

As a market research professional, it can be quite challenging to engage in discussions or online interactions with others in our field without encountering concerns about the potential impact of AI, particularly generative AI, on our industry and job prospects.

The conversation typically goes:

‘We’re all going to lose our jobs!’ 

‘We’ll get replaced by machines!’ 

‘How am I going to pay my mortgage?’

It does feel like we’re having a collective industry freakout.

However, having worked in market research in one form or another since the early 2000s, I feel it’s helpful to reflect on the technological changes and innovations I’ve seen arrive and whether these experiences can tell us anything about what to expect from the widespread arrival of AI. 

After nearly 20 years in the industry, what stands out is: 

1. Change is constant 

2. Previous technological change has felt less confronting, as it was always someone else’s job on the line. We, as researchers, remained somewhat shielded. 

This analysis not only looks into the past transformations but also highlights the unique factors that suggest this time around, we could be witnessing a truly distinctive shift.

Market analysis template

Save time, highlight crucial insights, and drive strategic decision-making

Use template

My first experience with market research was working as a CATI (Computer Assisted Telephone Interviewer) while completing my Masters in Organizational Psychology in 2002 at the University of Waikato (in New Zealand). 

For 2–3 hours, a couple of nights a week, I would hit the phones on behalf of the local newspaper (The Waikato Times), political parties, or consumer packaged goods (CPG) companies to conduct phone surveys. 

Depending on the client, we’d ask citizens’ views on local council issues (roads, water supply, etc.), understand their preferences for different types of yogurt or chocolate, or how they were voting in the next election.

It was a reasonably large operation, with 15–20 interviewers, mainly students, employed each night. 

Similarly, when I started my first job at TNS (now Kantar) in Auckland in 2005, their CATI had 50–60 interviewers at its busiest times. 

And, a few years later, when working at Ipsos MORI in the UK, they still retained a heavy face-to-face interviewer workforce who would intercept shoppers at Harrow on the Hill shopping center or go door to door, cold calling consumers and asking them to take part in a 40-minute (!) survey about magazine and newspaper readership. 

These days, it’s extremely rare to receive a call from a telephone interviewer. Dedicated CATI jobs are very limited, and the size of CATI operations everywhere has declined significantly. Nowadays, phone interviews are typically reserved for hard-to-reach audiences (older participants without internet access), niche B2B audiences, or political polling.

Having done both CATI and face-to-face survey roles early in my career, their gradual extinction didn’t feel like a loss. While there was good camaraderie in the CATI teams (especially when sharing stories of spectacularly rude survey rejection), it was hard to ignore the strong suspicion that you were unnecessarily annoying people trying to enjoy their evenings. The transition to opt-in online surveys felt better for the researcher and research participant.

When I began my second in-house market research agency role in 2007, our team of 13-15 researchers had two administrative staff (one solely looking after team travel) and two full-timers liaising with participant recruiters. 

One of the recruiter liaisons was a long-suffering Arsenal football fan who would take great pains to ensure we never ran evening focus groups on nights premier league games were on—especially with men ages 25–55, who would much prefer to stay at the pub watching an exciting match than turn up for an evening focus group. 

By 2010, all but one administrative staff was gone, the qualitative team looked after their own travel, and all respondent recruitment was outsourced to direct specialist providers. But the researchers were all untouched. Admin and ancillary roles were fair game for redundancy.

As a researcher at the time, you felt like a critical component of the business. It seemed like the only way to get fired as an agency market researcher was to be either thoroughly incompetent or actively criminal—helping yourself to the participant cash incentives, for instance. 

This era brought plenty of change within the individual researcher role, with many tools created to empower researchers (often at the expense of other roles). Historically, roles and processes (particularly in larger market research agencies), were incredibly fragmented.

In 2011, while working in London as part of a market insights team, we drafted all the surveys in Microsoft Word. Any changes were marked up and recorded before being sent to Bulgaria or Romania (essentially low-cost labor centers) to be scripted before launching online. 

If the agency client made any changes, these were marked up on the Word document and resent to Bulgaria or Romania. The survey script would be updated, new links created, and retested.

Clearly, a senior accountant in the business thought this was a good idea, and it must have kept many Bulgarian and Romanian programmers and scripters employed. Yet there were significant language barriers, and any changes to a questionnaire script meant multiple emails across opposite time zones, with inevitable errors and delays.

2016: Enter flexible survey tools

I noticed the advent and widespread adoption of survey scripting programs, such as Qualtrics, in 2016.

This development marked the start of individual researchers essentially doing it all themselves: designing a survey, launching it, and amending it once live. 

Outsourcing indeed still happens. However, the arrival of such tools immensely sped up the entire market research process.

Just as conducting market research has become simpler with the rise of easy-to-use tools, the analysis process has undergone similar changes, with fewer steps required to get insights and make sense of the data. 

Previously, once a survey collection was complete, I would take the raw data and send it to an analyst for cleaning. I’d also include a ‘tab spec’ of the subgroups to identify significant differences between age groups, sex, and socioeconomic brackets. 

I might also have specific requests to look at regression, correlation, MaxDiff (aka, Best-Worse ranking), etc. Any further changes to the analysis would involve contacting my analyst and recutting the data. 

Now, using data visualization tools like PowerBI, I do a lot of my own analysis—chipping away at the relevance of specialist analyst teams for anything other than truly specialized analysis—for instance, developing a customer segmentation from data analysis.

Introducing powerful, user-friendly scripting programs and sophisticated analysis tools has made life much easier for researchers. But just like when survey scripters became obsolete, these changes to the analysis process undoubtedly had downstream impacts on people performing market analysis. 

However, aside from the need to upskill on these new software products, these changes have only increased the importance of the researcher’s role.

Before the COVID-19 pandemic, there was a growing trend in conducting online qualitative interviews and focus groups, but there was still a strong demand for in-person sessions. I initially had reservations about online or phone interviews for research, doubting that these methods could provide the depth of insight or establish the same rapport with respondents that face-to-face interactions could achieve. In some cases, we even avoided phone interviews for these reasons. 

Nevertheless, since the pandemic hit, there’s been no looking back. The simplicity of scheduling and relative ease of use of platforms like Zoom and Teams have made it impossible to justify much face-to-face qualitative research. 

In addition, there are undoubtedly significant cost advantages of not hiring a viewing studio, paying a researcher to travel, or feeding participants and clients. 

Moreover, I don’t think any of my qualitative ex-colleagues are missing the 8:30 pm–10:30 pm focus group in a stuffy studio, with dry sandwiches and respondents who are either misrecruited or wishing they were elsewhere. 

Remote data analysis has also come to the fore. Shared analysis tools like Miro, which allow users to whiteboard their analysis from any location they want (without needing to decipher your director’s handwriting!), have dramatically changed how researchers collaborate. 

I regularly work on these platforms with Australian, Canadian, and US researchers from my Wellington, New Zealand home. Previously that would have required jumping on a long-haul flight, so these changes simultaneously save time, money, and carbon emissions.

So where does this leave us now? The pace of change in market research feels faster, driven especially by changes in work habits due to the pandemic. 

AI (as I interpret it) has been around for a while and creeping into research products, mostly like just another piece of tech. 

Social media monitoring and aggregating services that promise to track impressions and mentions of a brand, category, or its relevant attributes have been AI-powered for a while. 

Similarly, plugins like Beautiful AI use AI-powered presentation and slide design—an attractive prospect to a Gen Xer like me who has middling design skills.

Products like SignalAI, which tracks media mentions and conducts sentiment analysis with user-friendly, semi-automated reporting, are a timesaver and efficiency boon for researchers, helping them navigate and make sense of a sea of information.

Other tools like Otter.ai essentially remove the need for interviews to be manually transcribed, making life easier (and more profitable) for qualitative research specialists. 

To be fair, transcriptionists of yesteryear were amazing at what they did, turning around 30 A4 pages in 24 hours and using specialized setups with foot pedals to start and stop the audio recording. Still, almost overnight, another research-adjacent role, mostly filled by part-timers, is nearly obsolete.

AI-generated transcripts, in particular, are a significant time-saver for researchers. For example, there’s no longer a need to tediously review or rewind your audio or manually remove the inevitable ‘ums’ and ‘ahs’ in natural speech. 

The average transcript once cost anywhere between US $60.00 and $120.00, depending on the length and content. Comparatively, an annual licence for an AI transcription tool costs about as much as you’d have paid for a single transcript. 

Overall, the early introduction of AI across many digital products was a net benefit to most researchers and nothing to get overly concerned about.

I started hearing about ChatGPT, as many of you did, between late 2022 and early 2023. It was a novelty, with people using it for ad hoc tasks, like coming up with content marketing ideas, writing a song in the style of their favorite artist, or developing a new cocktail recipe with the leftover banana liqueur and creme-de-menthe in the liquor cabinet (okay, maybe that last one was just me). 

I assumed it would be similar to many other AI tools or tools claiming to have AI functionality—useful, potentially making my life as a researcher more efficient. And, as with all previous technological changes I’d experienced, I would remain the locus of expertise, providing insight and answers to my clients.

However, from hearing about other researchers' experiences and dabbling with it myself, it quickly became clear that this technology is different. Generative AI can deliver many value-added services that my colleagues and I essentially ‘sold’ to clients, more or less for free. 

According to OpenAI representatives(the creators of ChatGPT), at least 80% of the US workforce will have at least 10% of their tasks impacted, with 19% having 50% of their tasks impacted by large language learning models like ChatGPT. 

Survey researchers have been identified as potentially experiencing the second highest AI-related impact on their roles, right behind interpreters and translators. 

It’s worth pointing out that market research is hardly the only industry affected. Writers, content creators, and software developers, among many others, are experiencing massive changes within their industry due to tools like ChatGPT.

AI has quickly become ubiquitous within day-to-day market research. I am part of a protein snack start-up called Mīti (Māori for meat), which produces healthy beef jerky products. We wanted to investigate whether the Māori name had negative connotations across Asia as potential export locations. Instead of embarking on an arduous project with in-market research experts across the globe, we got a quick read via ChatGPT on whether the name had any significantly negative associations in various Asian languages.

Similarly, when I get market research briefs from clients, they have already done some initial digging via ChatGPT—using it as a starting point for their thinking or even writing their briefs using ChatGPT.

I suspect researchers will see increasingly fewer (or thinner) client briefs as AI gets smarter and more expansive.

Having watched so many colleagues in ancillary market research roles go over the years, it does feel scary that it might be the researcher’s turn to get replaced or at least downgraded in importance.

I touched on some examples of where AI is meshing into market research. In part two of this series, I go into more depth and explore the specific disruptions within the market research industry we’re likely to see over the next 3–5 years and where hope within the research industry remains.