Tanker operators embrace AI 

“AI can be our smart colleagues working quietly in the background, checking operations against real-time data and compliance rules… If done right, AI will not replace the human operator but work alongside us – 24/7, 365 – just like tanker operations.” 

That is how Maris Investments chief executive, Manish Singh, characterises the evolving relationship between shipboard officers and artificial intelligence (AI). Rather than envisioning a technological takeover, he describes AI as a junior colleague: attentive, precise and reliable, but still learning and always under human supervision. 

In recent years, AI tools have gained traction in areas such as emissions reporting, voyage optimisation, predictive maintenance and commercial fixture analysis. According to Mr Singh, the practical value lies in automation of routine compliance functions. “Several applications now auto-generate everything from shipboard logs to emissions reports and flag if you are about to breach any regulatory gaps,” he notes. “This allows officers to concentrate on navigation and safety-critical operations, rather than administrative tasks.” 

Although AI is no longer speculative in tanker shipping, its implementation remains uneven. BW Group vice-president, head of group IT and digitalisation, Patrik Desanti-Fettkenheuer, cautions that despite a surge of interest, most projects remain in pilot mode. “You might see a lot of initiatives going on,” he says. “But which of those initiatives really push the needle on revenue?” 

Mr Desanti-Fettkenheuer points to research showing that while four in five companies have tested AI in at least one function, fewer than one in five have achieved measurable business outcomes. “Below US$0.5M you will not have a real AI project,” he says. “Then you must account for legal advice under the EU AI Act and the additional burden of integrating with legacy platforms.” 

The issue of data quality is central. Tanker companies typically operate a suite of siloed systems for maintenance, navigation, fuel tracking and commercial performance, all developed at different times and often with conflicting data structures. Mr Desanti-Fettkenheuer advises against deploying machine learning on top of these systems without first unifying and cleaning the underlying data. “If you do not have good data, do not start the AI project,” he says. “The output will lead nowhere.” 

Mr Singh echoes this concern. In his view, partial data sets and fragmented logic are a source of systemic risk. “Even well-designed algorithms will hallucinate or deliver false confidence if fed with incomplete or poorly tagged data,” he warns. “This is how operators end up with faulty recommendations that appear trustworthy.” 

As the technology matures, responsibility remains a complex and unresolved issue. When asked who holds liability if an AI recommendation results in a safety incident, Mr Singh draws an analogy with pilotage. “Ultimately, the master is in command. The technology – like a pilot – is there to assist. But the responsibility never transfers.” 

This principle underlines the need for clarity and audit trails. Mr Desanti-Fettkenheuer believes decision-support systems must be designed with visibility into the recommendation process. “You cannot have a black box on board,” he says. “Operators must be able to trace back how a conclusion was reached and challenge it if necessary.” 

These reservations have not prevented early adopters from identifying value in specific operational areas. Predictive maintenance is one such field. By monitoring vibration, pressure and temperature data across critical systems, AI models can detect anomalies and anticipate failures. According to Mr Singh, this is especially useful in environments where technical staff rotate frequently. “You no longer rely solely on memory or logbooks. The system builds a digital history of the machinery,” he explains. 

However, the financial returns are not always straightforward. Mr Desanti-Fettkenheuer notes that most well-run tankers already report technical availability of 98%. “If an AI project raises that to 99%, you must ask whether the capital outlay delivers a measurable return,” he says. “It depends on your daily rate and the scale of the fleet.” 

More compelling is the case for voyage optimisation, ranked by many operators as the area with the greatest short-term impact. By integrating weather routeing with fuel consumption data and hull condition analysis, AI tools can suggest course and speed adjustments that lower emissions and improve scheduling accuracy. 

Compliance automation also ranked highly among operators prioritising future investment. Given the volume of regional, flag-state and IMO reporting requirements, software that can pre-screen data for errors and generate formatted submissions is in growing demand. In Mr Singh’s says, “This is not about replacing the officer – it is about augmenting the officer’s capability, just as we have done with ECDIS, GPS and emissions control systems.” 

This focus on support rather than substitution is reflected in attitudes to autonomy. Industry sentiment overwhelmingly favours advisory systems that offer context-aware suggestions, rather than automated decision-making. Mr Singh argues the best-performing tools function like vigilant colleagues, “handling hundreds of small, real-time observations the human eye might miss, but always under operator control.” 

Mr Desanti-Fettkenheuer believes the path forward lies in education as much as engineering. At BW Group, he has introduced internal sessions to demystify AI for superintendents and shipboard teams. “Technology becomes valuable only when it gives a role in our process,” he says. “AI is not an IT project. It is a business project.” 

For both, trust will determine the pace and scope of adoption. This includes not only trust in the output, but also trust in the developers, governance frameworks and system interoperability. Mr Singh calls for a broader industry conversation about standards, interfaces and ethical design. “We do not just need human-in-the-loop oversight,” he says. “We need human operators firmly in command, trained to understand when to override and when to accept.” 

From a governance perspective, there is also an urgent need for clarity around responsibility and insurance coverage. As Mr Singh puts it, “We are heading into a world where the ship, the shore and the terminal, will be more connected than ever. That will only succeed if our policies catch up with our software.” 

Looking ahead, both expect that AI will be embedded into a broader ship-shore ecosystem, not adopted in isolation. This transition will involve more than procurement – it will require rethinking how processes are built, who interprets recommendations, and where final responsibility sits. 

“The future is not about turning tankers into autonomous systems,” Mr Singh concludes. “It is about creating intelligent ecosystems where machine and human collaborate effectively. AI is the cadet: smart, fast, full of potential – but still under the supervision of a seasoned crew.”