Requirements Analysis for LLM-Integrated Applications: An SDLC Perspective

Abstract Requirements analysis establishes the foundation of the Software Development Lifecycle (SDLC) by providing a systematic approach to identify business needs, define software requirements, and evaluate key considerations before implementing a software application. Although the SDLC remains applicable to Large Language Model (LLM)-integrated applications, the probabilistic nature of LLMs introduces additional engineering considerations that are not explicitly addressed by traditional requirements analysis practices. This paper examines the impact of LLM integration on the requirements analysis phase of the SDLC. It compares traditional software application development with LLM-integrated application development and identifies the additional engineering considerations that arise during requirements analysis. The paper proposes a structured set of engineering considerations to support early engineering decision-making by organizing them into six categories: Business; Functional and Technical; Data and Information; Security, Privacy, and Governance; Operational; and Validation and Quality. Each category is accompanied by practical guiding questions to assist software engineers, architects, business analysts, and technical teams in evaluating whether and how LLM capabilities should be incorporated into software applications. The proposed set of engineering considerations complements existing requirements-engineering practices by extending traditional requirements analysis activities rather than replacing them. By identifying LLM-specific engineering considerations early in the SDLC, organizations can improve architectural decision-making, reduce implementation risks, and develop reliable, secure, maintainable, and business-focused LLM-integrated applications. Purpose The purpose of this conceptual scholarly work is to provide a practical guide for addressing a common challenge in the early planning of LLM-integrated applications. Organizations may be interested in adopting LLMs or integrating them into existing workflows but may still struggle to translate that interest into clear, reviewable, and feasible software requirements. This work helps organizations, businesses, and technical professionals develop a conceptual understanding of LLM-related requirements before actual implementation begins. It is not intended to replace existing software development workflows or propose a new methodology; rather, it supports early discussion and decision-making during requirements analysis. Keywords Requirements Analysis, Requirements Engineering, Software Development Lifecycle, SDLC, Large Language Models, LLM, Software Engineering, Software Architecture, Generative AI.