Executive Summary
Policy Goal: Transform ICAG assessment systems through AI-assisted hybrid architecture that reduces examiner workload by 40-50%, improves turnaround times, and maintains assessment quality by 2030, positioning Ghana as a regional leader in responsible professional education innovation. The Challenge: ICAG faces escalating assessment capacity pressures as Ghana’s expanding economy drives unprecedented demand for qualifid chartered accountants. Limited examiner pools, marker fatigue, inter-rater inconsistency,
relatively high assessment costs, and resource constraints strain traditional assessment models. Extended turnaround times delay candidate progression and affct Ghana’s competitiveness in regional accounting services markets. Key Finding: Ghana’s fist empirical study using July 2024 ICAG examinations evaluated four AI models (Claude 3.5, GPT-4, Perplexity, DeepSeek) across 27 scripts from nine subjects, generating 216 assessments benchmarked against human examiners. Claude 3.5 achieved mean absolute deviation of 4.1 points when provided marking schemes (46.1% improvement), substantially outperforming alternatives. Performance remained consistent across Foundation, Application, and Professional levels. However, systematic overscoring of weak candidates necessitates mandatory human verifiation for borderline and failing candidates.
Strategic Recommendation:
• Pilot Implementation: Launch pilot program in 2-3 subjects by Q2 2027 with Claude or equivalent LLM pre-marking and mandatory human verifiation for borderline scripts and 15-20% quality assurance sample.
• Infrastructure Redesign: Optimize all ICAG marking schemes with explicit criteria through senior examiner workshops.
• Governance Architecture: Establish AI Assessment Oversight Committee by January 2027 with Chief Examiner, Council representative, external AI expert, and Technical Review Panel.
• Capacity Transformation: Redesign examiner roles to AI supervisors through comprehensive CPD programs in hybrid assessment management.
