“Agentic AI for qual MR”

Gen AI is an excellent tool for the analysis of heaps of transcripts of focus groups or recordings of consumer conversations. Models, such as GPT-4, Mistral, or Llama 3.1, provide accurate, well-crafted summaries, extract the key topics from the conversations, bring up verbatim quota illustrating key arguments, and even design new products and services meeting consumer needs and wants expressed in the discussions to the point.

But what about COLLECTING qualitative consumer insight? A new generation of gen AI models can conduct straightforward in-depth interviews and probe for specificity with open-ended answers in online surveys.
These AI interviewing agents clearly fall short of the subtle skills of a trained moderator and obviously need input and training for each format and topic. However, with agentic AI we see a new field of “automated qual research” emerging, positioned somewhere between classic online surveys and human-moderated “true and deep” qual consumer insights.

data IQ has developed two demo prototypes that give an idea of how the leading model, GPT 4o, performs the task of “qualitative interviewing”. Please note, that these are experimental, early-stage prototypes. They are part of an extensive research project initiated by data-iq, exploring the potential and the best training/instructions for GPT and other LLMs.

1: A very simple agent that knows how to probe open-ended questions in a perfectly neutral, still assertive, and motivating way:

https://tinyurl.com/probeopenendeds

2: A slightly more sophisticated agent that conducts a straightforward, simplified in-depth interview based on a short discussion guide with a couple of questions:

https://tinyurl.com/gptinterview

We are looking forward to your feedback and thoughts.

contact us at: stefan.oglesby@data-iq.ch