RecoverAI 2030: An Explainable AI-Supported Project-Controls and Schedule-Recovery Framework for Saudi Infrastructure Programs
Year: 2026
Type: preprint
Abstract: Large infrastructure programs generate schedule updates, progress records, procurement status, design approvals, requests for information, changes, resource data and risk registers. Project teams may detect delay only after negative float and milestone slippage have become difficult to recover. This paper proposes RecoverAI 2030, an explainable decision-support architecture that converts ten transparent project-control signals into a recovery-pressure tier while preserving the baseline, critical-path logic and professional judgment. The criteria cover critical-path erosion, negative float, progress variance, productivity deterioration, design and approval lag, procurement delay, interface pressure, resource overload, schedule-quality deficiency and change exposure. The artifact is computationally stress-tested using 96 reproducible synthetic monthly schedule snapshots generated with fixed seed 2033; no Primavera P6 file, employer, client or live-project record is used. Scores ranged from 4.6 to 91.7, with a mean of 44.7. The model assigned 34 snapshots to monitoring, 31 to focused recovery, 22 to integrated recovery and 9 to executive intervention. Median tier stability was 100% under 1,500 weight perturbations per snapshot, while the least stable boundary case retained its tier in 61.3% of trials. A controlled approval, procurement and resource-improvement scenario reduced recovery pressure by 7.0 points on average. Artificially suppressing critical-path erosion and schedule-quality gaps understated pressure by 7.6 points. An illustrative recovery simulation reduced modeled finish slippage by 10.1 days on average, but this is a mathematical scenario rather than observed performance. RecoverAI 2030 contributes an auditable triage and action architecture, not an autonomous scheduling system. Field calibration, prospective validation, schedule-quality gates, data governance and human authorization are required before deployment.
Author: Ali, Qaisar Imtiaz
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7204492139
Cites:
Cited by:
Related to: