Agentic AI Control Tower for Offshore EPCI Marine Logistics Optimisation in Multi-Contractor Island Projects

Abstract

Abstract Offshore engineering, procurement, construction, and installation work depends on a marine logistics chain capable of balancing urgent construction demand against limited vessel capacity, field-entry and towage approvals, crew compliance, berth congestion, weather windows, fuel and freshwater availability, and cost responsibility across several projects operating on the same island. In many projects, the information needed to make these decisions is scattered across cargo manifests, vessel schedules, permit trackers, daily reports, emails, and spreadsheets. This paper sets out a practical agentic artificial-intelligence control-tower model that brings those fragmented inputs into one predictive and auditable decision environment. Deterministic safety and compliance gates are combined with bounded AI agents covering cargo prioritisation, voyage consolidation, vessel and tug assignment, permit and crew monitoring, weather and berth risk, bunkering forecasts, and project-level cost attribution. The prototype dashboard and workflow were assessed against an anonymised 12-week operating dataset from concurrent offshore construction projects on an ADNOC-operated island in Abu Dhabi. Performance under the observed baseline was compared with a model-assisted re-planning scenario. Average deck-space utilisation rose from 31% to 89%; combined landing-craft productivity increased from about one to 3.2 sailings per week; and completed cargo-barge towage cycles increased from about one to 3.8 cycles per active barge per month. Cargo readiness confirmed before berth booking improved from 57% to 93%, while average berth-related waiting fell from 11.2 to 3.1 hours for each affected voyage. Fuel-allocation completeness increased from 68% to 97%, and daily fuel-forecast error declined from 19.5% to 3.5%. The model also surfaced towage-approval, crew-rotation, weather, berth, and regional-disruption risks before mobilisation. Its main contribution is an integrated control-tower architecture that evaluates operational constraints and commercial ownership together, while preserving human authority over every safety-critical decision.