Dr. Lucy Tambudzai Chamba is an experienced lecturer, educational assessment specialist, researcher, and education consultant with over 20 years of experience in education, assessment, curriculum development, and management. She holds a DPhil in Management Sciences from Durban University of Technology, along with qualifications in Development Economics, Business Leadership, Business Studies, and Education. She currently serves as a Research Coordinator and Senior Manager at the Zimbabwe School Examinations Council (ZIMSEC) and is involved in postgraduate teaching and doctoral supervision. Her research interests include education, educational technology, entrepreneurship, public administration, and artificial intelligence in education. She has also co-developed an AI-based item-writing platform for national examinations and received the 2024 COMESA Excellence Award in Zimbabwe for the Education sector.

Toward Intelligent Assessment Ecosystems: LLM Orchestration and Multi-Layered Quality Gates for High-Stakes National Examinations The rapid digital transformation driven by Education 4.0 has exposed the limitations of traditional item-writing methods in national examinations and necessitates intelligent assessment systems that are scalable, transparent, and pedagogically aligned. This paper presents the design and prototyping of an AI‑enabled item generation and review ecosystem tailored for national assessments. The system leverages large language model (LLM) orchestration to extract topics and learning outcomes from syllabus documents and generate draft assessment items grounded in curriculum standards. It integrates a multi‑layered human quality‑gate workflow, First Examiner, Subject Matter Expert (SME), Head of Department (HOD), and Quality & Performance Management, ensuring that AI‑generated items undergo rigorous review for clarity, difficulty, accuracy, and curriculum alignment. The prototype incorporates role‑based access control (RBAC), detailed audit trails, analytics dashboards, and an emerging item‑bank architecture to support sustainable, enterprise‑grade assessment processes. A structured workflow (“generate → review → refine → approve → bank”) improves traceability, reduces authoring time, and enhances consistency across subjects. The system progressively improves item quality through a feedback‑memory mechanism that refines prompts based on reviewer insights, advancing toward a predictive and adaptive generation pipeline. A key contribution of this work is the incorporation of reliability, validity, and fairness safeguards throughout the item‑generation process workflow. Reliability is strengthened through standardized generation prompts, structured reviewer scoring, and consistency checks enabled by audit analytics. Content and construct validity are reinforced by grounding all generation processes in syllabus‑derived learning outcomes and by requiring multi‑stage expert verification. Fairness is promoted through controlled review by diverse subject experts, bias‑sensitive prompting, and blueprint‑driven coverage to avoid over‑representation or omission of learner subgroups or curricular domains. These mechanisms collectively ensure that AI‑assisted items maintain the psychometric integrity required for high‑stakes national assessments. Aligned with the Education 4.0 framework, this work contributes a practical architectural model for intelligent assessment ecosystems, demonstrating how AI can augment national examination processes for increased public value while preserving governance, transparency, and pedagogical integrity. The prototype illustrates pathways for integrating automation with human oversight to support SDG 4, strengthen institutional digital capacity, and scale high‑quality assessment delivery in rapidly evolving educational environments. Keywords: AI‑enabled assessment; Intelligent learning ecosystems; Large language model orchestration, public value