EXCLUSIVE: Bringing AI on to the bridge

Ardmore has spent the past two years building a partnership with the maritime AI and voyage optimisation company DeepSea to bring AI into the day-to-day voyage execution across our fleet.

In 2025, we began rolling out DeepSea’s Neuro device, an onboard hardware unit that sharply cuts the latency and data loss that traditionally slowed down digital tools. For an owner like us that is integrating new systems onto vessels that already carry their own sensors, communications setups, and reporting requirements, that reduction in latency is a very significant benefit that matters more in practice than any single optimisation algorithm.

While latency and data loss can sound like a back-office problem, the difference between a speed or routing recommendation reflecting current conditions compared to one reflecting conditions from twenty minutes ago can be significant on the bridge. An improved data pipeline with a reduced time lag gives our teams on the bridge and onshore a shared picture that is far closer to real-time and therefore far more operationally valuable. This is what makes AI-generated guidance worth acting on rather than something to second-guess.

In practice, that plays out as small, frequent adjustments rather than dramatic ones. This can look like slight speed reductions into forecast currents, or a course change made an hour earlier than a crew might otherwise have chosen. None of these decisions look remarkable individually. However, compounded across a voyage, and across a fleet, these small corrections are where significant fuel and emissions savings are found.

Through our engagement with the weather routing provider WetterWelt, every voyage is supported by data and by their experienced meteorologists. Safety is the overriding priority, but efficiency is a constant secondary objective. The aim is to reduce voyage time and fuel consumption where it is safe to do so and, after each weather-routed voyage, review performance and calculate the time and fuel saved. This allows us to quantify the financial and carbon impact on a voyage-by-voyage basis, rather than relying on fleet-wide averages. In 2025 alone, weather routing reduced Ardmore’s fleet emissions by 1,289 Mt of CO2, which is more than double the 546 Mt achieved in 2023.

An important and sometimes overlooked part of the process of adopting these technologies is building sufficient confidence in the data and recommendations to then act on it. A speed recommendation or routing adjustment is only useful if the officer receiving it believes it reflects real conditions, based on their experience, expertise, and observations. That’s why we treat training and onboard confidence as just as much a priority as the software integration itself.

None of this happens in a vacuum, and the technology is only ever half the story. Neuro’s rollout ran alongside drydockings, hull coating renewals, and biofuel trials across the same vessels, which all competed for the same limited windows of operational downtime and crew attention. In our experience, technology rarely fails on the spec sheet; it fails at the adoption stage. A rollout that goes well on one vessel because someone took the required time to properly explain it, listen to any concerns and answer questions might stall on another vessel if that step is skipped.

Ardmore’s fleet emissions fell 14% year-on-year in 2025, and that reduction reflects sustained work across the fleet. It also sits within a wider picture: trade patterns, port waiting times, and drydocking schedules all shape a year-on-year figure, which is why we look beneath the headline number as well as at it. What we focus on internally are the individual components – hardware like Neuro sitting alongside software, weather routing, and hull performance – managed as one portfolio rather than a bet on any single ‘silver bullet’. None of it counts for much, though, without a simple truth: no technology delivers anything until it earns the trust and confidence of the decision-makers on the bridge. That’s a human factor, and it’s a part of the technology rollout process we never take for granted.