The rapid development of artificial intelligence (AI) technologies, particularly Large Language Models (LLMs), in recent years has radically transformed organizational processes; however, this development has also created a complex management problem for decision-makers regarding how to identify the right project topic and select the most suitable model. The aim of this study is to provide an analytical decision support framework for identifying an AI project topic aligned with corporate strategies and selecting the most suitable LLM for that project. Within the scope of the study, an integrated hierarchical structure encompassing technical, financial, strategic, and ethical dimensions was developed by integrating current studies in the literature and expert opinions. The Analytic Hierarchy Process (AHP), a Multi-Criteria Decision Making method that enables the consistent quantification of multi-dimensional judgments in complex decision settings, was used to solve the decision problem. The analysis of the data obtained from five expert participants revealed that the “Security and Ethics” criterion had a much higher weight than the technical performance and cost criteria in both project selection (47%) and model selection (50%). When the alternatives were evaluated in the empirical application, the “Media and Content Management” project was determined as the most suitable investment due to its strategic alignment, and the “Llama (Meta)” model was selected as the most suitable technical infrastructure due to its data sovereignty and on-premises deployment capability. This study contributes to the literature by adding case-based evidence to an emerging consensus that AI investments are increasingly guided by a security- and ethics-oriented approach rather than by a cost-focused approach. Its main differentiator is the coupling of AI project selection and LLM selection decisions within a single AHP hierarchy.

Keywords

Large Language Models (LLMs); Analytic Hierarchy Process; Multi-Criteria Decision-Making; AI Project Selection; LLM Selection