An Intelligent Multi-Agent Framework for AI-Assisted Requirements Engineering Using Large Language Models
An Intelligent Multi-Agent Framework for AI-Assisted Requirements Engineering Using Large Language Models
Morteza Khani Dehnoi1
1) Department of Computer Engineering, Faculty of Electrical and Computer Engineering,Quchan University of Technology
IOI Code
:
ABEB-ABKCDA
Publication :
4th International Conference on Recent Advances in Engineering, Innovation and Technology - Belgium(eitconf.com)
Abstract :
Requirements Engineering (RE) is a fundamental phase of the Software Development Life Cycle (SDLC), significantly influencing software quality, development cost, project success, and stakeholder satisfaction. Despite decades of advancements in software engineering, the RE process continues to rely heavily on manual elicitation, analysis, validation, and documentation of stakeholder requirements. These manual activities are often labour-intensive, error-prone, and difficult to scale in modern software systems characterised by rapidly evolving requirements, distributed stakeholders, and increasingly complex architectures. While recent advances in Large Language Models (LLMs) have demonstrated remarkable capabilities in natural language understanding, semantic reasoning, and automated content generation, existing approaches primarily employ LLMs as standalone assistants, offering limited support for coordinating the interconnected activities that constitute the complete Requirements Engineering lifecycle.
This paper proposes an Intelligent Multi-Agent Framework for AI-Assisted Requirements Engineering that integrates Large Language Models with specialised autonomous AI agents to automate and coordinate key RE activities, including requirements elicitation, classification, ambiguity detection, conflict resolution, prioritisation, traceability management, validation, and Software Requirements Specification (SRS) generation. Unlike existing LLM-based approaches, the proposed architecture adopts a collaborative multi-agent paradigm in which specialised agents exchange information through a coordinated workflow, enabling continuous refinement and validation of requirements throughout the engineering process.
Keywords : requirements engineering large language models multi agent systems artificial intelligence software engineering