This study presents a framework for applying Large Language Models (LLMs) to software quality improvement, utilizing the iterative approach of the Deming Cycle. After applying a bibliographic method that integrates quantitative and qualitative analysis, current practices and essential indicators were analyzed, identifying important metrics such as task automation, reduced development time, improved error detection, and enhanced code quality. The framework defines a strategic combination of LLMs that overcome the shortcomings of conventional approaches such as the Software Development Life Cycle (SDLC) and agile methodologies, in terms of flexibility and accuracy. For the validation of the framework, expert surveys were conducted, and the results were analyzed using a Likert scale based on the IADOV technique to assess their perceptions. Complementary analyses were also conducted, along with a strategic assessment through a SWOT analysis and an economic assessment that included the return on investment (ROI). This analysis demonstrated a positive economic effect and highlighted the usefulness of the framework in demanding development environments. The results obtained establish the framework’s ability to adjust to complex and dynamic projects by driving constant improvements in software quality.