A Generative AI-Assisted Software Requirements Engineering Framework for Nigerian Enterprise Systems

Software requirements engineering in Nigerian enterprise system projects is frequently characterised by incomplete specification, stakeholder communication failures, and requirements volatility that contribute significantly to project overruns and delivery failure. This dissertation develops and validates a generative AI-assisted requirements engineering framework for Nigerian enterprise systems, contributing original research at the intersection of AI-software engineering and enterprise systems requirements literature. A design science methodology is employed across four artefact refinement cycles. Phase one reviews requirements engineering frameworks, large language model capabilities in software tasks, and enterprise systems deployment literature in African contexts. Phase two conducts empirical field studies at six Nigerian enterprise software projects in Abuja (ERP), Lagos (HR management), Kano (supply chain), Port Harcourt (hospital management), Enugu (government workflow), and Maiduguri (aid coordination) through stakeholder interviews and requirements document analysis. Phase three develops the GENREQ framework, which orchestrates GPT-4 Turbo for requirements ambiguity detection, consistency checking, domain-specific terminology normalisation for Nigerian regulatory language, and prototype user story generation from natural language stakeholder inputs. Phase four evaluates GENREQ through a controlled experiment involving 30 requirements engineers on standardised requirements elicitation tasks, measuring requirements completeness, ambiguity reduction, and stakeholder satisfaction. GENREQ reduces requirements defect density by 38 percent compared to unassisted requirements specification. The dissertation contributes GENREQ as a validated framework and advances generative AI application theory in requirements engineering. Keywords: generative AI, requirements engineering, enterprise systems, Nigeria, LLM