Natural language processing is the most obvious benefit of AI – the ability to understand what shoppers say in text messages, social media, open-ended survey responses, etc., is a great leap forward. It’s not as likely to kill jobs — but it may change the jobs that are needed. The risk is that people will believe what the system says and not check it/curate it/validate it. AI is just a program and is capable of producing the wrong answer.

Don’t just give up and say AI will produce something new, perhaps good or bad for consumer insights, for jobs — ARGH! I believe it’s worth a deep examination to see where AI can add value to the goals of research, i.e. for informed decisions with greater cost and time efficiencies. AI is a tool that can assist in data gathering and data analysis. I’m currently using AI to help reduce the time it takes to gather consumer insights by aggregating responses in real time during the course of an hour’s interview among a sample of 100 (or more) target respondents. It’s giving clients a chance to be on the front line with their consumers. That’s just one example. AI can add value to proven research procedures. And AI offers unique opportunities for new methodologies. If you think of it simply as “machine learning” you can lose the potential it offers because it can be intimidating. Instead consider the inroads AI can make in realizing broader and deeper consumer insights. It’s a context that will… Read more »

It’s easy to see how AI has already been applied to text interaction. The source data is massive, residing not in one-to-one conversations but billions of social media posts, reviews, and other online comments.

Quantitative phone research is another frontier. At first, artificial intelligence may run along side manual research gathering as machines are “trained” to do quantitative research. Customer service presents an illustrative model. Machine learning has automated routine customer support activities so that an increasing number of responses are handled by AI and a decreasing number of exceptions are escalated to a human representative. The same process can be used to automate quant phone surveys.

Over time, machines may revolutionize qualitative market research as well. Programmed branch questions will give way to dynamic questions assembled on the fly as machines hear “new“ responses and seek to probe further.

All of this frees up the researcher for higher level tasks around synthesizing rather than gathering results.

Repetitive tasks that are currently done by researchers are the easiest things to automate with AI. I think that will include survey design, fielding and analysis, particularly sentiment analysis and language processing.

The key parts of research that should never be replaced include:

• “What is the key question we need to answer?”;
• “What will we do with the findings when they are complete?”;
• “what are the implications of the findings for the brand and the business as a whole?”

Benefits include reducing time for a research project and therefore reducing cost. The risk is that researchers and marketers begin to rely on AI to answer questions that require some level of intuition and subtlety to find the real “nugget.”

I think, respectfully, these may be the wrong questions. At the risk of offending the two or three friends I have left in the marketing community, let me be intentionally provocative and play Devil’s Advocate here. A good prior question might be, “If AI proves to be an effective market research tool will traditional marketing jobs continue to exist outside the four walls of what are now client companies?” If these systems progress as their advocates suggest, why would you need a market research company in the first place? Why not just have a small in-house marketing capability and some really great systems? Marketers love to sell their, “value added,” analytics, but in my experience the data always speaks more clearly, and accurately, for itself. And if a smart system can identify patterns indicating potential problems or opportunities, help draft language for questions, target consumer micro-segments, gather data and provide fairly sophisticated analytical tools why invest in outside marketing? S, I don’t see AI killing jobs — per se — although a lot of low-level… Read more »

This statement is concerning: “three-quarters of the 250 market research decision-makers surveyed believe that the data produced through AI will be more accurate than it is today.”

There is no evidence, only promises, at this point to suggest that AI will make anything more accurate. My AI, in fact, is simply Big Data with a new name.

Is this an example of researchers with a “digital bias”? A inherent assumption that if it’s AI it must be better?

The technology we call AI today will do some good things. But like other tech trends in the past, it’s not truly AI and it is overhyped. Especially for its ability to interact with humans or pull the human out of data sets.

AI has great potential for tracking behaviors en masse and mining the data flows for insights. Social sentiment analysis comes to mind. So would conversational commerce interactions. Machine learning is already a foundation for price optimization systems — aren’t they really a form of behavioral research?

For survey research, there remains a need for a qualified professional to design the hypotheses, the target sample, and the questions. An AI could be useful for faster data analysis, but interpretation of results requires a human touch, at least so far.

Consider that data analyses — including measures of validity and error — used to require laborious manual calculations. Today, even experts use time-saving analytics software. If AI-based tools can help reduce human time and effort, it’s possible that one outcome may be more research conducted on more questions of managerial interest.

Overall, I think that’s a bullish picture for the market research profession.