Lifecycle for Artificial Intelligence: a systematic literature review

Artificial Intelligence (AI) systems are increasingly deployed in safety-critical and regulated settings, yet many initiatives fail to progress from proof-of-concept (PoC) to sustained operation. This gap is driven by the non-deterministic and data-dependent nature of AI, which requires continuous governance of datasets, models, evaluation evidence, deployment pipelines, and post-deployment monitoring. Although numerous AI/ML lifecycle models and process frameworks have been proposed, the literature remains fragmented, with inconsistent terminology and uneven coverage of operational and trustworthiness concerns.This manuscript conducted a Systematic Literature Review to consolidate and map recent AI/ML lifecycle frameworks (2020 onward). We identify, screen, and synthesize the relevant body of work and organize it into AI/ML lifecycle studies, lifecycle models, and industry-specific applications. The synthesis produces (i) an integrated landscape of lifecycle models; (ii) an explicit, stage-based taxonomy with consolidated operational definitions spanning business understanding, requirements, data work, modeling, deployment, monitoring, and usage/impact; and (iii) a cross-domain analysis of synergies and gaps, highlighting recurring deficiencies such as the translation gap between business goals and ML metrics, limited guidance for drift-aware monitoring under delayed ground truth, and weak operationalization of governance, fairness, and auditability.By making AI/ML lifecycle stages comparable across heterogeneous proposals and contrasting industry-specific emphases, the findings provide an evidence-based foundation to help scholars and practitioners move AI initiatives from PoC to sustained operation in applied, multidisciplinary, industry-facing R&D projects.