The boundless promise and fragile foundations of AI at sea

The boundless promise and fragile foundations of AI at sea
June 5, 2026 https://splash247.com/the-boundless-promise-and-fragile-foundations-of-a...

Everyone in shipmanagement is talking about artificial intelligence. Fewer are being honest about how far the industry still has to go. Another chapter from our new shipmanagement magazine which has proven to be a big hit at this week’s Posidonia.

From emissions reporting to autonomous watchkeeping, artificial intelligence is reshaping how shipmanagers run their businesses. But beneath the enthusiasm lies a set of uncomfortable truths about data quality, governance gaps and the enduring primacy of human judgement.

A spokesperson for Middle Eastern shipmanager Noatum CSM opens with a confession that many in the industry are reluctant to make publicly. “AI for me is not widely used in its entirety. I think we all think it’s here, and we like to talk about its use, but in all honesty it’s not here yet in the maritime industry.” His diagnosis of the primary obstacle is structural rather than technological.

“The biggest challenge we have as an industry before we implement the full force of AI is to unite the maritime industry in terms of standardisation with regards to technology. Once we solve that puzzle, I think you will really see how AI can transform.”

It is a candid starting point – and a useful corrective to the breathless enthusiasm that tends to dominate conference panels on the subject. The gap between what AI promises and what it is actually delivering in day-to-day shipmanagement operations is, by most accounts, still considerable. That does not mean the tools are not being deployed. They are. But the use cases are, for now, mostly at the margins of the business rather than at its operational core.

Where AI is actually being used

Across the industry’s leading managers, a reasonably consistent picture emerges of where AI is finding genuine traction. The common thread is decision support rather than decision making – tools that surface information faster, reduce administrative burden and improve the quality of human choices, rather than replacing the humans making them.

Niraj Nanda, chief commercial officer of Anglo-Eastern, describes his company’s approach as using AI and advanced analytics as “decision-support tools, instead of decision makers.” Within the company’s sustainability and performance services function, AI analyses historical and real-time voyage and performance data to support fuel efficiency, emissions monitoring and compliance planning. “These tools provide insights and forward-looking indicators that help teams compare scenarios and assess options,” Nanda says, “particularly in the context of increasingly complex environmental regulations. They do not replace operational decision-making, but they strengthen it by providing better information earlier.” Anglo-Eastern is also piloting AI applications to reduce administrative workload in areas such as document retrieval, technical referencing and first-level analysis – “supporting crews and shore teams by shortening turnaround times and reducing repetitive work.”

Vikas Trivedi, co-CEO of shipmanagement at Synergy Marine Group, offers a similar inventory of current applications: documentation, emissions reporting, maintenance planning, performance analytics, procurement intelligence and workflow prioritisation. But he describes the boundaries around those applications, saying: “AI is a co-pilot, not an autopilot. If we cannot audit it, explain it and control it, we should not operationalise it.” In a safety-critical industry, he argues, “poor data, hallucination and weak governance are not small risks, and AI cannot replace human judgement at the point of decision.”

Tim Ponath, CEO of NSB Group, takes the operational description further, describing AI agent deployment across company workflows as “the logical payoff for all the effort we’ve invested to be genuinely ready for this technological shift.” But he is equally sharp about where managers go wrong. “The biggest pitfall? Thinking simply throwing AI licences at your team and calling it integration. AI alone cannot work wonders. The magic lies in identifying high-impact use cases and doing the foundational groundwork first.”

The human data problem

One of the more thought-provoking contributions comes from Soma Sundar Gollakota, co-founder of behavioural risk management platform Bigyellowfish, who argues that the industry’s approach to AI data is missing its most important dimension entirely. “Most AI in maritime today is built on process and systems data – vessel performance, fuel consumption, maintenance cycles. That data is reliable, structured and relatively easy to collect.” But human data is a different matter. “It requires trust, ethical collection practices and full GDPR compliance. Most platforms haven’t solved that problem, so they avoid it.” The consequence, he argues, is an AI that cannot tell you anything meaningful about the people running the operation – and people, he says, are “where operational risk actually lives.” His company’s philosophy cuts through the complexity: “AI when it’s needed, human when it matters. The decisions it informs always stay with a human being. That’s non-negotiable.”

Vinay Gupta, managing director of Singapore manager Union Marine Management Services, takes the longest view of any contributor, saying current shortcomings are not pitfalls but “stepping stones towards a safer, more efficient and more sustainable shipping environment.” He sees AI-assisted navigation systems as an early step toward continuous watchkeeping support and, eventually, autonomous shipping. “The possibilities with AI are infinite, and it already offers solutions to several underlying problems affecting the industry today – from manpower shortages to the adoption of complex technologies by inadequately trained personnel.”

The garbage-in problem

The most forensic assessment of AI’s structural vulnerabilities in shipping comes from Manish Singh of Maris Investments, whose work with maritime companies gives him a clear view of where the foundations are weakest. “The maritime industry at large is structurally weak on data quality and cyber risk,” he says. “Our sector still runs on fragmented, inconsistently structured and often manually entered data, so the old rule ‘garbage in, garbage out’ applies with extra force.” He adds a further dimension that tends to be underplayed in industry discussions. “AI can also sound authoritative while being directionally wrong, especially when models are leaning on patterns from adjacent industries.” And the cyber threat compounds everything. “Reported maritime cyber incidents and attempted intrusions have surged in the last 12 to 18 months. You have a setting where enthusiasm for AI can easily run ahead of the governance and security disciplines needed to use it safely.”

Nanda at Anglo-Eastern identifies the same structural risk from a different angle. “The main pitfall at present is assuming that technology itself delivers outcomes. In a safety-critical industry, over-reliance on automated outputs without proper context, validation or oversight can introduce risk rather than reduce it. AI models are only as good as the data, assumptions and constraints behind them. Without experienced professionals to interpret outputs and understand operational realities, technology can create a false sense of certainty.”

People first, always

The final word belongs to Sebastian von Hardenberg, CEO of Bernhard Schulte Shipmanagement and president of InterManager, who distils the industry’s collective experience into a clear hierarchy of concerns. AI’s greatest pitfalls, he says, are “data quality, integration complexity and the risk of over-reliance on algorithmic outputs in safety-critical contexts.” But the solution is not primarily technical. “The success of AI in shipping depends less on the sophistication of the tools and more on how well we prepare and empower our people to use them. AI must enhance, not replace professional judgement in a risk-intensive industry like ours.”

It is a conclusion that runs through almost every contribution to this debate, expressed in different ways by managers of very different sizes and philosophies. The tools are here. The data foundations are not yet ready. And the humans – for now, and for the foreseeable future – remain indispensable.

This article was written by Claude

To access the full Splash Shipmanagement Market Report 2026, click here (https://issuu.com/sinoship/docs/shipmanagement_report_2026?fr=xKAE9_zMzMw)