ResearchPod Summary
Requirements elicitation interviews are critical for software success but are cognitively demanding, requiring interviewers to balance active listening, topic coverage, and adaptive questioning. Traditional training focuses on pre-interview preparation, leaving interviewers unsupported during the actual conversation. This study introduces an AI-assisted workflow consisting of two phases: offline script generation grounded in business goals (extracted from SEC 10-K filings) and online script management that provides real-time topic tracking and on-demand follow-up questions during Zoom interviews.
The researchers conducted a between-subjects quasi-experimental study with 43 interviewer-stakeholder dyads. One group (AI-assisted) used the AI workflow without prior formal elicitation training, while the other group (training-only) received one hour of traditional elicitation training and managed their interviews manually. The study evaluated three dimensions: script quality (using a best-practice rubric), interaction quality (measuring topic coverage and follow-up frequency), and the depth of elicited requirements (measured by the refinement of goal models).
The AI-assisted workflow significantly outperformed the training-only condition in several metrics. AI-generated scripts scored higher on a quality rubric (92.8 vs. 74.8 out of 100). During the interviews, AI-assisted participants covered fewer topics but asked significantly more follow-up questions per topic (3.43 vs. 1.15) and adhered more closely to their scripted questions (86% vs. 69%). Furthermore, the requirements artifacts produced by the AI-assisted group contained a higher fraction of low-level, actionable goals, suggesting that the AI support helped interviewers move beyond high-level abstractions to more concrete system requirements. Participants reported high utility for the topic-tracking feature, indicating that the AI successfully reduced the cognitive load associated with managing the interview flow.
This research demonstrates that AI can act as a scaffold for less-experienced interviewers, potentially democratizing the ability to conduct high-quality requirements elicitation. By shifting the focus from manual script management to adaptive probing, the AI-assisted workflow changes the trajectory of the interview, leading to more refined and detailed requirements. This suggests that future requirements engineering processes could rely more heavily on AI-integrated tools to bridge the gap between stakeholder needs and technical implementation.
AI-generated third-party summary by ResearchPod. Not official content or an endorsement by the paper authors or affiliated organizations.