ARTIFICIAL intelligence has dominated the maritime‑tech debate for the past 18 months, but industry leaders warn that the sector risks mistaking speed for progress.

Across Posidonia this week, shipowners, managers and class societies delivered a consistent message: AI is beginning to generate real operational value — but only for companies that have already built the digital foundations to use it safely and effectively.

The week opened with a blunt assessment from Diana Shipping director and chief executive Semiramis Paliou, who called AI “the elephant in the room” at the 10th Capital Link Maritime Leaders Summit – Greece. The industry, she said, still struggles to grasp the technology’s implications, yet AI will reshape shipping more profoundly over the next five years than geopolitics, fuel transition or any other headline issue.

“I’m convinced we’ll be seeing a different shipping industry in five years,” she said. “How, I’m not quite sure, but we need to open up this dialogue so we can start sharing information and knowledge.”

That tension — vast potential, limited understanding — ran through the week’s discussions.

Defining AI before deploying it

Andy McKeran, chief growth officer at Lloyd’s Register, urged the industry to stop treating AI as a single, catch‑all capability. Speaking at a Lloyd’s Register/OneOcean seminar, he noted that AI’s usefulness varies dramatically depending on the task.

He pointed to Lloyd’s Register’s work modernising legacy components in the OneOcean platform. Generative coding tools such as Claude, he said, have slashed development cycles: software modules once requiring two years to rebuild in VB.NET can now be completed in four days.

“The speed of that change is enormous,” he said.

But acceleration brings risk. Companies must protect intellectual property, deploy AI only in secure environments, and ensure strong governance and clean data. Without those foundations, AI becomes a liability rather than an enabler.

AI, he stressed, is an additive capability — not a replacement for fundamentals. Without connectivity, governance and structured processes, it is “potentially useless”. With them, it becomes a powerful augmentation layer. “It’s not the silver bullet.”

LLMs, knowledge management and the rise of shadow AI

Maria Kartalou, head of planning at Thenamaris, highlighted the immediate value of large language models in document search and knowledge management. Safety management systems, class circulars, inspection reports and technical manuals can overwhelm crews and office staff.

“You have to spend sometimes hours to find the one specific process that you need to follow,” she said. LLMs, she argued, can surface answers instantly — in any language — and preserve operational knowledge that currently sits buried in superintendent emails.

But she echoed the warnings: without high‑quality data, LLMs can confidently deliver wrong answers. Operators must learn to challenge AI outputs.

For Matthew Maheras, president of AMMITEC and CIO of Metrostar Management Corp, the bigger threat is “shadow AI”: employees using free, unvetted tools without oversight. A captain optimising a voyage with a consumer LLM or a chartering manager pasting a charter party into a public model could expose sensitive data or rely on flawed information, potentially breaching contracts or invalidating insurance.

“Shadow AI is being used by users without us knowing, or without central orchestration,” he said. Governance frameworks such as ISO 42001, he argued, are urgently needed.

Shipowners invest as class societies urge caution

Shipowners at the Capital Link summit expressed optimism tempered by pragmatism.

Seanergy Maritime Holdings chief executive Stamatis Tsantanis said his company has already made two major AI investments — one in technical management, another in psychological assessment tools for seafarers — and called the progress “spectacular”.

Safe Bulkers CEO Polys Hajioannou said his company is preparing to adopt AI once internal evaluations conclude, with the biggest gains likely in automating repetitive office tasks. “If you need one hour now to complete paperwork for fixing a vessel, and AI can produce it in two minutes, you will use AI,” he said.

DryDel Shipping chief executive Costas Delaportas pointed to benefits such as live voyage P&L monitoring, but warned that system upgrades can exceed $1m. “We need to pay because there is no other solution as the market goes,” he said.

Class societies reinforced the need for strong foundations.

Bureau Veritas Marine & Offshore president Alex Gregg‑Smith said AI is maturing quickly but must remain transparent and human‑supervised. Above all, it should “make the seafarer’s life better”.

DNV Maritime interim CEO Cristina Saenz de Santa Maria noted that class has used AI for years, but LLMs have accelerated the conversation. She warned that as IT and OT systems converge poor governance or weak data quality could make AI detrimental. Some companies, she added, are too consumed by geopolitical pressures — such as navigating the Strait of Hormuz — to prioritise AI adoption right now.

Beyond point solutions

McKeran also cautioned that many companies are still chasing isolated AI use cases rather than addressing systemic inefficiencies. Fleet performance analytics often attract disproportionate attention despite limited data volumes.

“You may find that there’s actually more operational efficiency and process improvement and communication improvement between ship and shore than there is potentially analysing a single voyage,” he said.

He also warned of rising costs as AI providers consolidate market power. Subscription models costing around $200 per user per month could increase sharply. “We’ve seen… where they suddenly realised that they’ve got a hold on the market, the price has gone up by five times.”

The foundations will determine the outcome

Across Posidonia, the message was clear: AI’s potential in maritime is enormous, but the industry is not yet ready to harness it safely at scale. Technology is advancing faster than governance, understanding and data quality.

Shipping’s AI moment has arrived. Whether it becomes a breakthrough or a burden will depend entirely on the foundations laid today.