Splash247: Has maritime AI already become a commodity?

Published by Splash247

As artificial intelligence becomes increasingly commoditised, maritime software vendors face an uncomfortable question: if everyone has access to the same underlying models, how can their products stand apart? SplashTech canvasses the industry and discovers that the next competitive battleground has remarkably little to do with AI itself.

There is a growing irony at the heart of shipping’s artificial intelligence revolution. The technology promising to transform maritime software could also make competing products increasingly difficult to distinguish.

As vendors rush to incorporate large language models into everything from voyage optimisation to fleet maintenance, access to sophisticated AI is no longer particularly exclusive. So what happens when everyone is effectively using the same technology?

“The model is not the product,” argues maritime technology commentator Kris Vedat. “A vendor who thinks plugging in an AI model is the strategy will look identical to everyone else – because it is everyone else’s strategy too.”

His assessment is echoed across an industry increasingly convinced that the competitive advantage is shifting away from the technology itself towards proprietary data, maritime expertise and the ability to solve genuine operational problems.

Christine Sullivan, chief product officer at Veson Nautical, argues that the differentiator is the combination of human expertise and validated information. For AI to deliver meaningful recommendations, she says, it must understand the complexities of voyage economics, port operations, vessel lifecycles and commercial fixtures rather than merely process generic information. That places considerable value on the accumulated knowledge held by established maritime software businesses.

Mikko Kuosa, CEO of NAPA, points to the company’s calculation engine, refined over decades, as an example of technology that cannot simply be replicated by plugging into an AI model.

“A generic AI model has no idea what safe cargo distribution looks like on a specific roro class, or what a class society will accept in a stability calculation,” he says.

NAPA’s AI-powered Permit to Work dashboard, used by Virgin Voyages and Ritz-Carlton Yacht Collection, illustrates the distinction. Users can interrogate fleetwide safety information through natural language, but the intelligence depends on the structured operational data and safety logic beneath the interface.

The same principle applies to navigation. Yarden Gross, CEO of Orca AI, says a generic model cannot reliably distinguish an unlit fishing boat from wave clutter in darkness without extensive maritime-specific training. His company’s systems have learned from more than 200m nautical miles of navigation data.

“The model is becoming the easy part,” Gross says.

For smaller specialist vendors, however, this emphasis on proprietary information is also an opportunity. Antonis Malaxianakis, founder and CEO of HarborLab, points to six years of curated port data supporting the company’s work in disbursement account management. Understanding the peculiarities of port tariffs, invoices and local operating practices provides a degree of specialisation that generic AI tools cannot readily reproduce.

Damien Bertin, business director at Opsealog, makes a similar argument about vessel performance. His company’s benchmarks draw on more than a decade of information from over 1,000 offshore support vessels, allowing its analysis to distinguish between operating modes and vessel types. Across ADNOC Logistics & Services’ offshore fleet, a programme combining data analytics with marine efficiency expertise has reported fuel savings exceeding 20,400 cu m. For Bertin, measurable results ultimately matter more than the sophistication of the underlying algorithm.

However, data alone is not sufficient. Aristos Philis, CEO of KeelX, asks whether an AI platform actually understands a customer’s vessels, knows what happened yesterday and can take meaningful action.

“Can it actually do something, or can it only answer questions?” he asks. That distinction between an intelligent interface and an operational tool may become decisive.

At SpecTec, customer success director Tom Davy highlights spare parts forecasting as an example. AI can analyse maintenance histories and predict future requirements, but the commercial benefit only materialises when those predictions feed into procurement decisions.

Sven Brooks, CEO of ScanReach, believes another differentiator is understanding why crews ignore alerts and how information becomes a practical decision.

“The model is becoming the least interesting part of the stack,” he says.

The question also exposes a potential disadvantage for longstanding software suppliers. Years of accumulated data offer a competitive advantage, but ageing architectures and cumbersome interfaces can make it harder to exploit new capabilities.

Jacob Iversen, European business development director at Weathernews, argues that cloud-based, AI-native systems may be better positioned to connect operational data and analytical tools, although they still need specialist knowledge.

Meanwhile, Petter Andersen, vice-president of shipping at StormGeo, compares AI models to engines: fitting the same engine to two different cars does not make the vehicles perform identically. 

Related Posts