AI and machine learning in the shipping sector: from possible to practical

The 2026 Offshore Support Journal Conference Asia addressed a wide range of issues, challenges and opportunities facing the offshore support vessel sector. Among them, inevitably, was the potential benefits of digitalisation and AI in the maritime space, a subject that was addressed in a presentation by Rolls-Royce Marine Automation Ship Systems manager Kanza Nassabi entitled, ‘From possible to practical: closing the gap between AI capability and operational value.’ 

In her presentation, Ms Nassabi said applying machine learning on a vessel involves a number of questions. The first is, can the technology perform the task we want it to? Can it collect the data and can we detect patterns? “The answer in so many cases is, yes, we can,” she told delegates. The next question is, can we trust the technology, she said. “Even if we collect the data and can detect patterns, are we ready to rely on it to make decisions and use it operationally?” And can we act on the insights that it provides? 

“On one side, we have the technology, the sensors, connectivity, the cloud and AI. This side of the equation is moving incredibly fast,” Ms Nassabi told delegates. “And then we have the other side, which is operations, where we have real people, real workflows, and real responsibility and, most importantly, real consequences. We can build incredibly sophisticated systems, but if we don’t understand who we are building them for, then we may be solving the wrong problem.

“Imagine that data seems to show a change in the condition of a cylinder in an engine. Detecting a fault is, potentially, very helpful, but if as a result we trigger an inspection of the engine earlier that was actually necessary, we might not find anything meaningful. And that inspection takes time, and it has a cost and, potentially, it disrupts operations.

“The system needs to give us the right information, at the right time, with enough confidence to act,” she said. “False alarms and unnecessary inspections lead to a loss of trust. If damage is detected too late, this results in more downtime and even higher costs. So when we talk about predictive technologies, earlier is not always better. The real value lies not just in identifying change as early as possible, but in identifying the moment when that information becomes actionable.

“A signal can tell us that something has changed, but it doesn’t tell us the whole story,” she concluded. “This is where human context comes into the picture. The algorithm sees the pattern, and the expert understands what surrounds it. 

“AI detects, but human expertise gives it meaning. Together, they support better informed decision-making. Using machine learning and AI isn’t about removing the expert from the process, it’s about giving him a better point from which to start.”