AI Is a Paradigm Shift
October 1st, 2025
At the ARF/MSI panel on synthetic data I was asked by Oded Netzer, (Professor of Marketing at Columbia Business School) where we will be in 5-10 years. My simple answer was that AI is a paradigm shift.
What we are witnessing is the beginning. Some of us look at the phone-web transition in the early 2000s as an analogy. While there are similarities, it’s an incomplete analogy. Data collection modalities have shifted over the last century from in-person to mail to phone to web. But the process of research hasn’t changed a whole lot even though many new analytical techniques have been implemented over time.
But AI is very different.
A direct way in which AI changes the game is through its infusion into the entire research process. LLMs are so capable that not using them for brainstorming, questionnaire development, analysis/report writing etc., is a sure sign of underestimating their potential. They can not only speed up every element of the research process but can also produce superior value and insights – when researchers use them as teammates and not just tools.
While that is great, it’s not a paradigm shift.
The greatest use of AI in research is through the development and usage of synthetic data. One simple way to think about it is through conjoint analysis. Imagine a conjoint study without and with a simulator. The former is good, but the latter is great. Now imagine having something like a simulator at every stage of the research process to generate ideas, run scenarios to test hypotheses, develop alternatives, etc. It can change the very fabric of research.
This is all true both in the quantitative and qualitative research domains. In quantitative research, for example, digital twins can help us test, refine, and shape ideas as a precursor such that valuable human data can be put to its best use. In qualitative research, digital personas can help marketers dive deeper into consumer needs and motivations with longform nuanced answers before getting to valuable human conversations.
But all this paradigm-shifting wonder comes with caution.
The technology of LLMs is so dazzling that some companies are being fooled into short-term exploitation. The title of my panel at ARF/MSI was “Pros and Cons of Synthetic Data.” I played with the words to suggest that it is more Con in the short term but more Pro in the long term – because, while the potential is enormous over the long term, there is a lot of hard work needed to get there.
Just trying to make a buck today truly will be a short-term con game.
