Generative artificial intelligence has rapidly entered enterprise software environments, transforming how organizations process information, interact with data, and automate knowledge-intensive tasks. While early applications have focused on content generation, summarization, and recommendation, the deeper architectural challenge lies in integrating generative AI into systems where outputs influence operational decisions. Enterprise software does not merely consume information; it executes actions, enforces policies, and produces outcomes that require reliability, traceability, and accountability. This paper examines the architectural transition from traditional data pipelines to decision intelligence systems. It argues that generative AI should not be treated as a standalone decision-maker, but as a probabilistic intelligence component embedded within governed decision pipelines. To address this challenge, the study introduces the concept of a Decision Materialization Layer, an architectural layer that transforms AI-generated outputs into structured, validated, explainable, and auditable decisions. The proposed model reframes generative AI outputs as intermediate signals rather than final answers. These signals are enriched with contextual data, evaluated against business rules, cross-validated when necessary, and converted into decisions that enterprise systems can safely act upon. Through this approach, organizations can bridge the gap between probabilistic AI behavior and deterministic enterprise requirements. By developing a framework for integrating generative AI into enterprise architectures, this paper contributes to the emerging field of decision intelligence systems. It offers a structured pathway for designing AI-enabled software environments that are not only intelligent, but also governable, observable, and dependable.
ILKER KANATLI (Tue,) studied this question.