Many software systems no longer meet their current and future requirements, exhibiting the need to be modernized. Software modernization is within the rapidly evolving landscape of software systems a highly relevant topic. Any modernization initiative involves a series of consequential decisions, most notably the selection of an appropriate method. Academic literature offers a substantial body of work addressing software modernization and decision making in this area, however the industry perspective remains underrepresented within decision making. In practice, formalized decision processes are rarely employed. Instead, intuition is a predominant factor in decision making, contrasting with the concept of rational, traceable and ethical decisions. This work aims to bridge the gap between academic literature and industrial practice concerning decision making in software modernization. Exploratory literature reviews on recommender systems and predictive analysis are conducted, complemented by semistructured interviews and a questionnaire distributed to industry participants. A combination deliberately chosen to approach the topic from both perspectives simultaneously. Drawing on these sources, a multi-step decision process is developed that integrates literature findings with current industrial practices. A notable finding is the significant discrepancy in terminology between the literature and industry. While the literature maintains a clear distinction between modernization strategies and individual methods, industry practitioners employ these terms interchangeably, as evidenced by the interviews conducted. The work further examines and discusses the applicability of recommender systems and predictive analysis within this process, and identifies and discusses the informational requirements and decision factors necessary to support both approaches. The properties of rational, traceable, and ethically decision making can be achieved through decision and estimation models, including recommender systems and predictive analysis. Recommender systems reduce the influence of intuition and human bias, while predictive analysis spans from intuitive approaches employed in software modernization to formalized techniques including machine learning, successfully applied in fields such as economics and medicine. The emergence of Large Language Models (LLMs) and AI-driven decision support systems presents an opportunity to combine both approaches into an integrated solution. To this end, an architectural design is proposed incorporating large language models, retrieval-augmented generation, and a feedback loop for continuous learning, enabling more rational and traceable decision making in industry while remaining accommodating of current intuition-driven practices. In doing so, this work addresses the absence of a structured decision framework in industry through the proposed decision process, and counters the reliance on intuition through the combined application of recommender systems and predictive analysis facilitated by LLMs.