Does generative AI reduce the mental effort required to gather and define software requirements?

Host: Welcome to A.I.S. Insights — powered by Living Knowledge. I’m Anna Ivy Summers. Today we are looking at a study titled "Requirements Engineering with Generative AI: Transformation of Requirements Elicitation and Cognitive Load," presented at the International Conference on Information Systems, or ICIS, in 2026, by researchers Adrianna Weronika Pawińska, Kevin Kintzel, Till Schirrmeister, and Georg Voronin. Joining me is our analyst, Alex Ian Sutherland. Alex, what is the core focus of this study?

Expert: Hello, Anna. This study investigates how generative AI influences requirements elicitation—the early stage where software teams discover, interpret, and define system requirements—and how integrating these tools reshapes the cognitive load experienced by practitioners.

Host: Most research and discussion on generative AI in software focuses on coding, such as code generation or debugging. Why is examining requirements elicitation important?

Expert: Requirements elicitation is an ambiguous, knowledge-intensive activity. Analysts must explore problem domains with stakeholders, uncover tacit needs, reconcile conflicting goals, and evaluate feasible solutions. Errors made at this early stage often cascade through the entire development lifecycle, leading to project delays or budget overruns. While organizations often assume generative AI will uniformly reduce mental effort, there has been limited evidence on whether it actually reduces cognitive burden or simply introduces new complexities.

Host: How did the researchers approach this question?

Expert: They conducted a qualitative case study at a European IT service provider that is currently integrating generative AI into its requirements engineering workflows. The authors carried out twenty semi-structured interviews with experienced practitioners and analyzed the data using grounded theory methods, interpreting the results through the framework of Cognitive Load Theory.

Host: What did the study find regarding how mental effort changes?

Expert: The central finding is that generative AI does not produce a simple, uniform reduction in cognitive load. Instead, it redistributes mental effort across different types of load. Automation reduces extraneous cognitive load by handling repetitive, administrative tasks such as transcribing meetings, summarizing discussions, drafting initial user stories, and formatting documentation.

Host: But new demands emerge to counterbalance those savings?

Expert: Yes. The study found that new cognitive burdens arise from the need to formulate effective prompts, evaluate tool capabilities, and systematically verify outputs to detect hallucinations or inaccuracies. Additionally, when stakeholders use generative AI to prepare their own documentation, analysts must invest extra mental effort to separate genuine stakeholder needs from generic model outputs. Among the twenty participants, eleven experienced a net reduction in cognitive load, nine felt their overall mental load remained unchanged, and one reported an increase in net cognitive burden.

Host: How does this interaction influence creative thinking during elicitation?

Expert: The study found that creative outcomes depend on the mode of engagement. When practitioners use generative AI reflectively as a sparring partner to test assumptions, ask assurance questions, and explore alternative perspectives, it supports creative problem-solving. However, when practitioners adopt outputs superficially without critical evaluation, it risks reducing original thinking. Furthermore, experience levels matter significantly. Junior staff often lack the domain knowledge required to critically assess AI-generated material, which in turn increases the oversight and review burden on senior practitioners.

Host: What are the main takeaways here for business leaders and technology teams?

Expert: Organizations should not assess generative AI adoption solely through time saved on initial drafts. Offloading routine documentation creates downstream demands for verification, contextual judgment, and system monitoring. To capture value safely, organizations need to train teams in prompt formulation and critical evaluation, ensure domain expertise remains central to the review process, and adjust workload expectations to account for these shifted responsibilities.

Host: In short, generative AI transforms the nature of requirements work from drafting toward critical verification and judgment. Alex, thank you for sharing these insights. And thank you to our listeners for joining us on A.I.S. Insights — powered by Living Knowledge.