Internal audit and defensible AI tender evaluation in UAE construction procurement systems
TL;DR
Cross-regime results show portability of the framework, but different emphases: Abu Dhabi places greater weight on criteria-aligned explanations, whereas federal and Dubai contexts depend more on reconstruction-ready audit trails.
Abstract
This study examines how artificial intelligence (AI)-assisted tender evaluation can be supported by governance conditions associated with legal defensibility in UAE public construction procurement, with emphasis on evidence required for oversight, supplier grievances, and audit. A comparative design covers federal, Abu Dhabi, and Dubai procurement regimes. Procurement defensibility is modelled as an organisational outcome produced by three governance mechanisms – explainability quality, bias-audit rigour, and audit-trail completeness (ATC) – with internal audit involvement (IAI) specified as an assurance enabler. Survey data from 330 procurement, internal audit, and legal/contract professionals are analysed using partial least squares structural equation modelling, with measurement invariance of composite models and multi-group analysis. All three mechanisms have positive, significant effects on defensibility, with ATC the strongest driver. IAI positively but modestly strengthens the audit-trail -> defensibility relationship, indicating that assurance routines increase contestability-by-design. The model explains substantial variance in defensibility (R2 = 0.62) and links defensibility to higher confidence in value-for-money and compliance (R2 = 0.30). Cross-regime results show portability of the framework, but different emphases: Abu Dhabi places greater weight on criteria-aligned explanations, whereas federal and Dubai contexts depend more on reconstruction-ready audit trails. The study offers a practical defensibility-readiness roadmap for AI-enabled tender evaluation.