From challenges to metrics: An LLM-driven DevOps recommendation system grounded in evidence-based mappings

DevOps is widely adopted, yet many organizations continue to face barriers such as integration gaps, unclear objectives, and fragmented decision support. These barriers limit the alignment of DevOps initiatives with organizational goals. This study introduces an empirically grounded recommendation system that connects DevOps adoption challenges to success factors, operational processes, and performance metrics. The system is built on a structured mapping model and uses a semantic recommendation engine enhanced with large language model capabilities. It interprets natural-language descriptions of challenges and generates context-aware recommendations by combining semantic retrieval with evidence from 378 peer-reviewed studies. Evaluation shows that the system outperforms keyword-based retrieval, achieving higher precision at rank three and improved expert-rated relevance. By linking practitioner challenges with systematically curated evidence, the approach provides actionable and auditable guidance for DevOps transformation planning. The results highlight the potential of semantic reasoning and structured knowledge extraction to support evidence-based decision-making in software engineering practice.