Splash247: Shipping’s AI arms race moves beyond the chatbot

Published by Splash247

As artificial intelligence becomes increasingly commoditised, maritime technology providers are scrambling to distinguish themselves. Industry leaders tell SplashTech why proprietary data, deep sector expertise and real-world results matter far more than the AI models themselves.

If every maritime software supplier can access the same artificial intelligence models, what exactly will make one platform better than another? For the executives responding to SplashTech’s latest survey, the answer has little to do with the cleverness of the chatbot.

The consensus is that AI itself is rapidly becoming a commodity. The competitive advantage lies in the information feeding it, the maritime expertise interpreting its output and whether it can make life easier for people at sea and ashore.

“The model is becoming infrastructure. What you build with it is the product,” says Aristos Philis, chief executive of KeelX. He argues that the questions worth asking are whether the technology understands a customer’s vessels, knows what happened yesterday and can take action rather than simply answer questions.

Christine Sullivan, chief product officer at Veson Nautical, puts validated data and human expertise at the heart of the proposition. The essential foundation, she says, is a secure, interoperable system of record with structured, real-time information. Making tedious tasks faster is useful, but AI’s bigger prize is stronger decisions about voyage economics, fixtures and port operations.

For some vendors, the differentiator is decades of information that a newcomer cannot quickly recreate. Opsealog’s business director Damien Bertin points to more than 10 years of performance data from over 1,000 offshore support vessels. He says this allows comparisons to reflect specific ship types and operating modes, rather than treating all offshore vessels alike.

In an ADNOC Logistics & Services programme covering more than 100 offshore support vessels, Opsealog reports fuel savings exceeding 20,400 cu m. Bertin stresses that data needs to be paired with people who understand operations if it is to deliver measurable results.

HarborLab founder Antonis Malaxianakis makes a similar case for six years of curated port information. That gives the company a basis for training AI to handle disbursement accounts, invoices and tariffs, where a plausible generic answer may still be an expensive wrong one.

SpecTec’s Tom Davy cites a more prosaic differentiator: 40 years of maintenance expertise behind AMOS. AI can forecast spare-parts demand, he says, but it only becomes valuable when procurement teams can act before a missing component causes downtime.

Out on the bridge, the consequences can be more serious. Yarden Gross, chief executive of Orca AI, says a generic system cannot reliably distinguish an unlit wooden fishing boat from wave clutter at 3am. His company draws on more than 200m nautical miles of navigation data. In an 828 nm feeder voyage assessed by Lloyd’s Register, it reports computer-vision precision of 94% and recall of 98.6% across 739 targets.

“Officers don’t want another screen to babysit,” Gross says. “The vendors that stand apart will be the ones built around how a watch is really kept.”

Hrvoje Mihovilović of ELNAV.AI pushes the safety argument further. Recognising a spoken helm order is one thing, he explains; confirming that rudder movement matches the instruction demands vessel data, timing and an understanding of the manoeuvre. He wants buyers to test how systems cope when inputs are missing or the model is wrong, rather than trusting a polished demonstration.

There is also disagreement over what constitutes a useful outcome. Symeon Chatzigeorgiou, strategy chief at DeepSea Technologies, argues that software offering recommendations alone leaves a gap between advice and action. The lasting edge, he contends, will come from linking AI to specialised hardware capable of direct, real-time control.

Mikko Kuosa, chief executive of NAPA, sees the distinction in a different place: the company’s stability calculation engine, accumulated operational data and workflows. Its AI-powered permit-to-work dashboard lets fleet managers interrogate risk information in ordinary language, but he says the value comes from the safety logic and structured records underneath.

Weathernews’ Jacob Iversen questions whether all platforms will ever use identical models. For weather routing, he says, years of observations, vessel performance reports and experienced meteorologists remain vital. As autonomous AI agents spread, buyers should insist on understanding and auditing how recommendations are reached.

Related Posts