DeepSea & Nabtesco — vessel fuel optimisation

What is established

Nabtesco’s 4 April 2025 release reports 3.4% fuel improvement on Malaysia Grace with TELEGRAPH AGENT and DeepSea’s Cassandra. Its stated method used ten days of alternating three-hour on/off periods at constant speed. This is a supplier-reported combined-system trial, not an AI-only or fleet-wide estimate.

Primary source ↗

The operating problem

A ship operator needs lower consumption without losing schedule reliability or compromising safety. The useful question is how a recommendation performs on a particular vessel under defined voyage conditions.

Illustrative evaluation workflow by AIStrategy.gr; not a disclosure of the organisation’s internal architecture.

Design a comparable pilot

Agree a vessel-specific comparison plan before deployment. Record displacement, weather, speed, route, hull condition and operational changes. Keep master and engineering authority explicit and log overrides.

What to measure

• Fuel per transport work under matched conditions
• Arrival and schedule performance
• Override frequency and reasons
• Measurement uncertainty and excluded periods
• Total integration, support and operating cost

What the evidence cannot settle

A short alternating trial may not represent seasons, routes or the full operating envelope. Check how weather and loading were treated and whether observed gains persist after deployment. Savings from a combined system cannot be assigned to one component without further evidence.

Questions for the project owner

• What is the baseline, comparison period and representative operating envelope?
• Which data are missing, and how do those gaps affect decisions?
• Who can override the system, stop the trial or correct an error?
• What results remain after integration, review, training and support costs?

Scale, revise or stop

Agree thresholds before the trial. Expand only when the responsible team can reproduce useful outcomes, explain remaining errors and sustain the operating process. Revise the design when benefits depend on exclusions that do not reflect normal work.

Editorial boundary. The primary-source account above is distinct from our proposed pilot design. Reported outcomes have not been independently audited by AIStrategy.gr.