Requirements elicitation for AI-Enabled process automation: A comparative review of Agile, Waterfall, and Hybrid implementation approaches
TL;DR
The defensible prescription is neither agility nor hybridity as a blanket label but a gate-differentiated lifecycle: plan-driven discipline where late discovery is irreversible, iterative discipline where early commitment is guesswork.
Abstract
Organizations adopting AI-enabled process automation are routinely advised to choose a delivery lifecycle that is agile, plan-driven, or hybrid as though the choice applied uniformly across a whole specification effort. That framing is mis-drawn. Building on the author's antecedent five-gate method for converting an ambiguous business problem into an AI-ready specification (problem framing, decision decomposition, data grounding, behavioural specification, and oversight and acceptance), it argues that the gates impose opposite lifecycle requirements. Gates resolving uncertainty about what the business wants reward short cycles, working artefacts, and rapid feedback; gates establishing fixed conditions reward front-loaded rigour, sign-off, and documentation that survives audit. The defensible prescription is therefore neither agility nor hybridity as a blanket label but a gate-differentiated lifecycle: plan-driven discipline where late discovery is irreversible, iterative discipline where early commitment is guesswork. Because empirical work already reports hybrid delivery as the industrial norm, the contribution is a basis for which hybrid, not an argument for hybridity. It is a structured comparative synthesis of published reviews and empirical work in requirements elicitation, agile and hybrid development, process automation, and process mining, read alongside applied literature of varying provenance; it is not a new systematic review and reports no primary data. Its outputs are a three-archetype classification of automation, an eight-dimension comparison matrix across three lifecycle ideal types, a mapping of elicitation techniques, including process mining, to gates and archetypes, a four-factor contingency model, a survey of how lifecycle pressures appear across applied domains, and eight propositions for testing. Keywords: Requirements Elicitation, Intelligent Process Automation, Agile Requirements Engineering, Software Development Lifecycle, Contingency Model, Process Mining, AI Governance.
