The State of AI in the Insights Industry

October 16, 2024

The market research world is changing at the pace of AI. How are you keeping up as more and more capabilities evolve?

The State of AI in the Insights Industry

It’s difficult to consume any content lately that doesn’t have a hot take on AI: what it really is, how it’s influencing any given industry, and what it all means for the future, especially the future of work. The Insights industry is no exception. While a comprehensive overview of everything AI currently touches is beyond the scope of this particular piece, we do want to explore the current state of things in general. Any new technology brings with it potential benefits, a slate of challenges, and many possible futures. At Knit, we want to be part of building a future where technology acts as an assistant to the real expert: you, the researcher.

Understanding AI in the Insights Industry

What we call “artificial intelligence” today isn’t an actual reproduction of human intelligence, as anyone who has used any of the more basic models can tell. Most of the models are LLMs- large language models- machine learning algorithms trained on large datasets. Generative AI like ChatGPT works as a kind of probability or prediction machine, producing in response to a prompt a distillation of what it has been trained on. It doesn’t think for itself, but it can help a tired human brain reorganize information or think through it differently.

How it’s currently being employed in the industry

AI is being weaved into every step of the research process, from how surveys are generated to how data is collected, processed, interpreted and more. We see it used now in:

  • Survey/questionnaire generation
  • Data collection and processing
  • Data analysis and interpretation (including sentiment analysis)
  • Personalization and consumer insights

The technology is better in some areas than others; organizing and recombining data is the easiest for it to take over and take off the plate of a human researcher. In its current iteration, it’s also better at setting a researcher up for deeper analysis than to be left to that task on its own. Sentiment analysis, for example, is notoriously difficult for robots to grasp since they struggle with sarcasm and other inexact forms of human communication.

Benefits of AI in Insights

Done right, AI can enhance the accuracy and efficiency of data processing, particularly at scale. Research professionals no longer have to spend hours hand-coding and processing data, freeing them up to do the kind of deep, interesting thinking that leads to more extensive and actionable consumer insights. The 2024 GRIT Insights Report highlighted which tasks insights professionals are automating with AI:

GRIT AI chart

In their own analysis, the GRIT team broke down the results this way:

“At the risk of oversimplifying, analysis seems to be the low hanging fruit of automation, possibly because of the expertise and computational capacity it requires and possibly because these solutions are more mature.”

Adoption rates differed between the segments they surveyed, emphasizing the different concerns and approaches of each. One of the major advantages of AI often highlighted by proponents is the potential in cost savings. Only a few segments agreed on this- technology providers and data and analytics- so the key to driving adoption of more AI automation might be emphasizing this potential benefit. Fewer segments reported using it for things like refining insights into presentations teams can bring to key stakeholders, saving time as a resource by getting the green light on projects faster. This difference could be attributed to distrust in the technology’s ability to do certain tasks well, a lack of awareness that it’s an option, or not having the technology available (due to lack of budget or other factors, like lack of buy-in from those same decision makers). As the technology improves, it can get better at predictive capabilities, like trend analysis. How much better it gets depends on the data it’s trained on, however.

Challenges and Limitations of AI

As with any new technology, AI comes with plenty of challenges and limitations– many of which get glossed over when it comes to enthusiastic discussions of its capabilities.

Data privacy and other ethical considerations

Any organization using AI needs to think critically about responsible management of consumer data. If it’s going into the AI they use, will it get deleted after processing? If it’s being stored, how is it being stored, where, and for how long? There should be complete transparency into all of these considerations for any potential customer to explore, plus the option to opt out at any time. Bias is unavoidable in anything created by humans and that includes the data AI is trained on. Teams need to be careful to control for bias in training data as much as possible to ensure that generative AI doesn’t parrot the same harmful stereotypes often found on the worst corners of the Internet. (Trying to avoid this by using synthetic data comes with its own set of issues, which we’ll touch on in another section.) There are other ethical considerations for organizations and individuals to take into account when tapping into AI, too. This technology uses a huge amount of resources, impacting the environment. Training the technology to be able to identify what humans consider inappropriate also means using human labor to do that training– and its often underpaid and psychologically damaging work. As an industry, there’s a need to develop and maintain ethical guidelines for AI use in market research that addresses all of these areas. At Knit, we’re dedicated to weaving these considerations into everything we do– ensuring our partners are set up for success in their work with us, to the highest standards possible.

How Knit is integrating AI

At Knit, AI is not just a tool; we’ve built our platform around it to enhance the efficacy and efficiency of the work you do. We envision AI as the ideal co-pilot for insights professionals, able to step in and streamline at every step of the research process, from study design to reporting– when wanted. In other words, the researcher is driving the work and the AI functions as a research assistant, capable of lightening the load of tedious tasks and expanding the speed and scale of the work your team is able to do. The goal is opening your time up to do more of the deep work that you want to do and less of the work you don’t. The AI can remix and present data in a way that helps you pull out the threads you need to put together those deeper insights. It can’t think for you, but with you. Our AI allows you to tap into the assistance you want at every step:

  • Study Design: Our AI assists researchers in crafting surveys based on their research objectives and other context.
  • Data Analysis: Our system intelligently analyzes quantitative and qualitative data, including complex open-ended text and video feedback, to uncover meaningful insights for researchers in minutes.
  • Quality Checks: AI-driven algorithms are employed to scrutinize the data for bad actors, fraud, accuracy and consistency, ensuring that the responses are high quality.
  • Report Generation: The AI synthesizes the analyzed data into comprehensive reports, creating narratives and summaries that are insightful and easy to comprehend.

Researcher-driven AI sets you up for success.

Want to know more about how Knit is applying AI? Reach out to our team for more details!

Technical challenges

The other set of challenges around AI are technical in nature and largely based around the enormous amounts of data needed to train AI models. With generative AI in particular, it’s only as good as the data it is trained on. Scraping the internet for data leads to issues with data quality, questions around copyright, and other complicated legal questions that are only starting to surface with more widespread use of the technology. Synthetic data is one proposed solution which aims to erase issues with data privacy and the ongoing need for new data, but it comes with its own set of challenges and issues. Some researchers have likened it to mad cow:

“Seismic advances in generative AI algorithms for imagery, text, and other data types has led to the temptation to use synthetic data to train next-generation models. Our primary conclusion across all scenarios is that without enough fresh real data in each generation of an autophagous loop, future generative models are doomed to have their quality (precision) or diversity (recall) progressively decrease. We term this condition Model Autophagy Disorder (MAD), making an analogy to mad cow disease.”

This could create a feedback loop where each generation becomes a lesser version of the one before, the further it strays from the richness of real-world data.

The Future of AI

Gazing into our crystal ball, the future remains cloudy but our stance does not: At Knit, we want to see researchers remain firmly in the driver’s seat with technology acting as a perfect droid copilot helping to navigate the process. The goal is to have a personalized research assistant who learns from the team it works with and is able to help bring more complex insights to light, at scale. It can’t replace you, because it can’t think like a researcher does. It’s there to lean on as a tireless support system. That’s how Condé Nast is using it through Knit as our CEO Aneesh Dhawan and Maja Benedict, Associate Director of Global Custom Insights discussed at Quirks Chicago in March of this year:

“It’s finding efficiencies with our team. We’re a lean, mean team so we have to use our time very wisely… us to focus our time on refining.”

They’re able to jump right in and see the findings right away, giving them more time to tell the deeper stories found in the insights. If that’s something it sounds like you and your team could use, our team is ready to help you get set up!

Get Started