Two uses of AI are already practical in research: querying a completed study and choosing a follow-up question for a thin open-ended answer. Both work against study data and instructions supplied by the research team, and both produce output that a researcher can review.
Querying Data & Commentary
Inside a controlled data environment, a prompt-based interface can query structured and unstructured study data. For example, researchers can:
- Ask specific questions about quantitative data (e.g., “What is the demographic profile of consumers most open to Product X?”).
- Request familiar cuts without building each table manually (e.g., “Show the distribution for Question 5 by generation among respondents in the South.”).
- Apply a defined framework to qualitative material and review the resulting themes.
- Search a controlled source library for relevant examples or prior work.
The output still requires review. A fluent answer can contain a bad calculation, omit an important exception, or rely on a source that does not belong in the analysis.
Probing Open-Ended Responses
Online surveys lost much of the back-and-forth probing that interviewers once handled in person or by telephone.
Generative AI can restore some of that probing by selecting a relevant follow-up question from the participant’s response and the study’s approved instructions. If a participant writes, “I like the product features,” for example, the survey might ask, “Which features did you like?”
This can add useful detail in product, concept, positioning, and advertising studies. The probe logic, respondent experience, privacy handling, and resulting data still need to be designed and reviewed by researchers.
The next article examines the more experimental use of AI personas and synthetic data.