Software quality evaluation has traditionally relied on manually defined metrics, expert judgment, and testing activities performed at selected stages of the software development life cycle. Although these techniques remain valuable, they often provide limited support for identifying complex quality patterns across large and heterogeneous software repositories. Artificial intelligence (AI), combined with data-driven analysis, enables organizations to evaluate software quality continuously by learning from source code, development history, testing records, issue reports, and operational data. This paper presents a research framework for data-driven software quality evaluation using AI. The framework integrates data collection, preprocessing, feature engineering, machine learning, quality prediction, explainability, and continuous feedback. It considers functional correctness, reliability, maintainability, security, performance, and usability as interconnected quality dimensions. The paper discusses suitable data sources, machine learning techniques, evaluation metrics, implementation challenges, ethical considerations, and threats to validity. It also proposes a reference architecture for integrating AI-based quality assessment into modern software engineering workflows. The analysis indicates that AI can improve early defect detection, prioritize testing activities, estimate maintainability risks, and support evidence-based quality management. However, successful adoption depends on data quality, model transparency, domain adaptation, human oversight, and continuous monitoring.