Why AI is an opportunity for the Insights industry as long as the human factor remains

Like many other industries, Insights is drowning in a glut of posts, articles, presentations and conferences (including a very good one just before the holidays from the ASC: link) about the current and potential impact of AI. With the plethora of perspectives, scenarios and forecasts, it can be daunting to decide on the direction to take, but it boils down to whether you’re a pessimist or an optimist. At SA Market Insights, we’re very much in the latter camp. If AI helps us get a better understanding of context when considering a client’s business issue, great; if it helps us design, set up and deliver client projects more efficiently and quickly, also great. And if we can uncover new perspectives, generate innovative ideas and deliver insights faster for our clients, even better. Here’s the areas where we at SA Market Insights think it can make a difference:

• Understanding wider business context and identifying study inputs: quickly identify industry dynamics that may be impacting our client’s business objectives and use these to inform study design and recommendations, e.g. competitor activity, technological change and regulatory updates.
• Enhancing study design quality: identify wider sets of response options, especially when lacking qualitative data inputs, ensure consistency, accuracy and tone of voice to improve data quality and reduce the risk of bias.
• Live data cleaning: identify erroneous responses, disengaged respondents and survey bot activity while fieldwork is live to improve data quality across the entire sample and reduce time spent checking.
• Multiple options for open text analysis: fast analysis of open-ended responses to enable a range of outputs: sentiment analysis, thematic analysis, messaging ideas, new product ideas, which will also provide more depth to quantitative data.
• Rapid visualisation: quickly generate charts, graphs, images and other visual effects from data and text for use in online reporting tools, charts and other deliverables.

This is exciting because from our perspective, we can deliver deeper insights to our clients faster than before. However, there are still a number of important considerations which mean we cannot use LLMs universally or without human oversight:

• Privacy and data security is paramount in our industry, therefore any LLM we use with client data must be private and trained on our own data, rather than a public domain tool.
• Clients are developing their own guidelines and rules about the use of LLMs both internally and by the agencies they use. These need to be accounted for when considering study design and how data and information is used.
• There are limitations with LLMs for use with less widely spoken languages. Translating text into English before inputting into an LLM may miss important inferences and nuances.
• Many aspects of research require a balance of both art and science, from survey design to the advanced analytical modules we use in our studies (which already incorporate elements of AI). AI is just not there yet when it comes to this balance, and the expertise and insight of a human remains critical.
• There is still enormous value to be derived from qualitative research, and the body language, tone of voice and inferences it allows us to observe for which there is no substitute.
• AI does a great job in some areas but is weaker in others, and the inputs used have a huge impact on the output – understanding these nuances is critical to successful implementation.

For all the benefits AI offers, we must always remember that our primary responsibility to our clients is to convey the opinions, needs and behaviours of their target markets and users in a compelling way that allows them to make decisions. This relies on our capabilities as storytellers and our acumen as advisors. Only with these, along with the rapidly emerging capabilities of AI and the LLMs it powers, can we offer something truly impactful to the clients we serve.