Could a Large Language Model Contribute Significantly to Requirements Analysis?

This research-in-progress paper presents a quasi-experiment in which three different ChatGPT-4 prompts (for system structure, analysis, and recommendations) are applied in standard or augmented form to the work system in each of three case studies (automated warehouses, ride hailing platforms, and medication administration systems). The augmented forms (treatments) are based on different sets of ideas. Each case study comprises 3000+ words. The prompts are detailed requests for responses of up to 500 words related to three steps (system structure, analysis, and recommendations) related to those cases. A null treatment serving as a quasi-control uses standard prompts for each case without augmentation. The first actual treatment is a revision of an analysis template used by MBA and EMBA students; the other six are sets of questions based on activity theory, a BPM design space, system principles, and three other approaches The research questions are whether ChatGPT-4 can produce a useful first cut at system structure, analysis, and recommendations and whether various augmentations of ChatGPT-4 prompts improve or extend the outputs significantly.