<p>This article examines how two English proficiency tests—the General English Proficiency Test (GEPT) and the BEST Test of English Proficiency (BESTEP)—have evolved from traditional gatekeeping instruments into learning partners. Anchored in the principles of Learning-Oriented Assessment (LOA) and the glocalization imperative, both tests illustrate how domestically developed assessment systems balance local educational priorities with international standards of test quality and fairness. Central to this evolution has been the adoption of the socio-cognitive framework, which provides a foundation for construct definition, test design, and empirical validation, enhancing validity and reinforcing stakeholder trust. The GEPT, launched in 2000, was designed as a level-based examination aligned with major educational stages to support lifelong learning. To promote positive washback and enrich score interpretation, the GEPT conducted Common European Framework of Reference for Languages (CEFR) linking studies and integrated technological innovations to enhance score reports with diagnostic information, learning suggestions, and personalized insights into learners’ strengths and weaknesses, fostering autonomy and engagement. Drawing on this foundation, the BESTEP was introduced in 2023 under the Ministry of Education (MOE)’s Program on Bilingual Education for Students in College (BEST program), which seeks to strengthen bilingual education in higher education. From the outset, BESTEP integrated technology-enhanced feedback, adaptive resources, and cross-disciplinary collaboration, benefiting from innovations developed through GEPT projects. Its design and validation, guided by the socio-cognitive framework, ensure contextual relevance and international comparability. By linking curriculum and assessment, BESTEP contributes to the evolution of context-specific assessment practices, supporting pedagogy while meeting policy goals. The transformation illustrates how language assessment can be understood through the lens of glocalization: local policy responsiveness, learner-centered design, and contextualized tasks are combined with frameworks and validation evidence that ensure international recognition. Still, challenges remain, including balancing accountability with learning impact, upholding fairness and accessibility in AI-based services, and sustaining collaboration across test developers, educators, and engineers. Additional complexities arise in deploying AI scoring models and maintaining their long-term operational validity. By reflecting on these challenges, the article identifies pathways for sustaining innovation while safeguarding validity, fairness, and educational impact. The evolving practices of GEPT and BESTEP demonstrate how a local testing organization can reimagine assessment as a catalyst for learning, dialogue, and reform. More broadly, the adaptive experience offers insights into the evolution of language testing across Asia, where local-global integration continues to shape new trajectories of practice.</p>