.png)
“Generative AI is going to be a market researcher’s superpower.”—Xander Jefferson, Product Manager at Harris Quest
The media buzz around generative AI in the workplace has far from flattened over the last nine months or so, and across the board, investments in technology are accelerating.
But most of the headlines highlight the most obvious—and splashy—AI capabilities.
You can create song covers by artists who are no longer alive, dream up quirky scenarios to be “painted” in the style of Picasso, or write your kid’s English paper on the eve of its due date.
But how is generative AI developing in the market research space?
Harris Quest, a new suite of research tools is pushing to weave generative AI into its offerings, from DIY survey generation to insight summaries, and more. I sat down with Harris Quest’s product manager Xander Jefferson, and Gary Topiol, CEO of Maru/HUB (creator of one of Harris Quest’s new products), to learn more about where their teams are investing in generative AI and how researchers should think about bringing it into their workflows.
The opportunities are endless. But both present measured perspectives on the limits of the technology—and the enduring importance of market researchers to herald the integrity of the data they’re scoping for.
“I think that generative AI is going to be a market researcher’s superpower,” Jefferson said at the start of our conversation. “It’s going to enhance what their current capabilities are.”
How?
By increasing speed and efficiency and lowering cost.
Jefferson recommends starting with a quick win when it comes to getting your team on board: Let AI summarize key findings from focus groups.
“AI is really good at ingesting this type of data. You can output the key themes or bullet points from open-ended survey data. You’re no longer having to do costly translations of these particular types of datasets or have a human read through hundreds—or sometimes thousands—of lines of open-ended comments just to extract key themes from a data set. AI can do this, and in lickety split time compared to what a human can do, at the fraction of the cost.”
What else can generative AI do for researchers?
Me to Google: How much data does the average enterprise collect over a year?
Google to me: Tons.
And even though leaders agree that data is crucial to their success, 55% of organizations report that their data is “dark” (i.e. not discoverable). It’s also been found that workers spend an average of 1.8 hours a day looking for information they can’t find.
“Within many enterprises and agencies, you have these huge tombs of knowledge that are partitioned into products or projects and hidden away,” Topiol says, “and there’s no real way to mine value out.”
This is where generative AI can make a huge impact on efficiency and discoverability. Gen AI can make it possible to query huge amounts of data using natural language. Think of it like a private ChatGPT: You can ask a chatbot questions about your company’s repository of data, and the AI will spit back insights, findings, or citations from the full library. Harris Quest is doing this with QuestAI.
This not only is a great use case for market researchers who want more well-rounded access to their data vault, it can be used across the company for improved data discoverability.
For market researchers, generative AI can be a huge time saver when it comes to actually drafting surveys.
With Harris Quest, Topiol is adding this feature to a do-it-yourself surveying tool, QuestDIY. Now, instead of spending time drafting new questions or copy-pasting them over from previous surveys, you provide the kind of survey you want to create (brand awareness, concept testing, campaign tracking, etc.), the number of questions you’d like in your survey, and a brief summary of what you’d like to see, and generative AI uses the inputs to create a survey for you.
“You still have to have some understanding of research best practice,” Topiol advises. “I can review something quickly that maybe previously would have taken me an hour or two to draft and another hour to script. I can jump through all of that and spend time tidying up the survey and making sure it’s as good as possible.”
Data query isn’t the only value generative AI can bring to data discovery and analysis—it can also summarize key findings from data sets.
At The Harris Poll, Jefferson is focused on developing QuestAI, a tool that can ingest open-ended data sets like focus groups, podcasts, and interviews and create a summary of what was found.
Not only does this save a huge amount of time for the researchers who would have to listen to and interpret hours of focus group data, it also can bring more of a data-centric approach to that analysis.
I asked Jefferson about what a machine might miss when it comes to the nuances of market researcher. Does missing tone of voice or body language impact the depth of the findings?
While the moderator should still be a key player in the discussions that follow, Jefferson says that AI does a great job quantifying the discussion. Maybe the moderator felt like a certain participant was obsessed with diesel trucks in a focus group on electric vehicles. AI would be able to pick up that it wasn’t as big a part of the conversation as it seemed or that it actually only came up once and took up a small percentage of the conversation.
“You can use AI for tonality and the words they’re using to understand what the sentiment is. You start to put a quantified picture together of the keywords or the brands or the questions that invoked emotional responses,” he says.
Up next is the idea of using synthetic data to lower the cost of surveying, particularly with hard to reach audiences.
Running market research with traditional methodologies can be expensive. If your company is looking to gain insights around your consumer, it could cost you $20,000 to $50,000 on average.
What if we could create lookalike audiences from huge amounts of historical data to anticipate how certain personas would respond to a survey?
“It’s essentially a way of relating data without going through the expense of asking anyone anything,” Topiol puts it simply. “There are some very good use cases for synthetic data, including building data sets to train analytical models.”
While this might lead to fairly generic findings, according to Topiol, it could give brands a head start or an early read to optimize their paid for research. This technology is very much a work in progress, but the potential is really exciting.
If you give a mouse a cookie, he’s probably going to ask for a glass of milk.
And if you give a researcher data, she is probably going to want a data visualization.
Generative AI can help with that too.
Historically, researchers have relied on tools like Tableau or Microsoft Power BI to bring their data to life in bar graphs or pie charts. But Jefferson is keen on the future of data visualizations run by generative AI.
“What I’m seeing in the space is the ability to use natural language to say, ‘Show me a visualization of the percentage of people in the US who drive electric cars in bar chart form.’ If you plug that into the ecosystem of things, you can pair more of that quantitative and qualitative analysis.”
It’s not perfect yet, and certainly isn’t a replacement for data scientists. But it can support with number crunching functions and basic market research analytics.
“You’re automating a process where you don’t need to use an extensive, slow, clunky software to get the outputs then drop them into PowerPoint,” says Jefferson.
If you’re nervous about bringing generative AI into your workflow, Jefferson suggests to start off by just letting these tools check your work.
Go through with whatever process you already have in place for data analysis, but also see what AI might pull out from your source. You might get an extra insight or two—without really lifting a finger.
“You’re expanding the scope of the quality of your own research on top of what you’re already delivering to the client,” he says.
As for Topiol, he advises to look at what the real value is versus the hype.
“I believe we risk just making it incredibly easy to do bad quality research through these tools, but we don’t want to do bad research. We need to think about how we use these capabilities to do more efficient, insightful, high-quality research. There’s a lot of noise, and there’s a lot of potential, but if I’m going to use these tools today, how can I use them to really improve what I am currently doing?”
Start with low stakes experiments, think critically about the business value that’s on the table, and let the technology help you become more efficient.
So what’s next? Jefferson and Topiol definitely have an idea of where these tools are going: Analysis and extrapolation of structured data, data democratization and personalization, and globalization of AI tools to remove biases are just a couple of those ideas.
But ultimately, the market researcher isn’t going anywhere.
“The role of the market researcher is always going to be required” according to Jefferson. “This is someone who understands a client’s business, their needs, the project requirements. They’re the enablers.”
With no shortage of use cases for researchers to take advantage of, their work is going to become richer with the value brought by generative AI. Decades of dusty data may be brushed off and put to good use, research will become cheaper and more accessible to conduct, and hours of time savings will relieve researchers from tedious manual work.
