ELNAV.AI: Artificial intelligence should assist the bridge, not audit it

A master mariner with more than 30 years at sea across cargo ships, ferries, and cruise vessels, Mihovilović founded ELNAV.AI to build practical AI safety systems for bridge operations, with a focus on watchkeeping, human-factor risk, and real-world deployment.In line with his company’s mission, he sees the real breakthrough happening when shipping stops treating AI as a feature label and starts treating it as part of a safety management system. “That means clear limits, traceable outputs, proper change control, degraded-mode behaviour, human-factors testing, and evidence from real operating conditions,” Mihovilović says.

Shipping understands that AI will matter, he tells Maritime CEO, but says the governance around it is less developed, especially for safety-critical AI, which can’t be just seen as software.    “That distinction matters. AI can support safer watchkeeping, earlier risk recognition, fatigue-aware intervention, and better learning from weak signals,” Mihovilović explains. “But if AI is designed mainly for remote monitoring, scoring or blame-by-replay, it can damage trust and weaken just culture.”

Mihovilović tells Maritime CEO that his position is not anti-AI and that he is more focused on a pro-safety, pro-seafarer view.   “Shipping should use AI where it improves safety and reduces harm. But we should resist designs that quietly turn professional seafarers into data sources for remote scoring and post-event blame. The principle I keep coming back to is simple: assist the bridge, don’t audit it,” he states.

Shipping is already willing to pay for certain AI solutions to alleviate pressure from fuel savings, emissions regulations, insurance pressure, crew availability and regulatory complexity. However, Mihovilović sees the industry as still being very pragmatic about it.

“Owners will pay when the business case is clear, the operational burden is low, and the system does not create new problems for the crew. The strongest case is not ‘buy AI because AI is modern’, the stronger case is: reduce operational risk, reduce workload, improve compliance, preserve professional competence and give management better evidence before something goes wrong,” he says.

The ELNAV.AI founder admits a certain ‘fear of missing out’ factor being present in AI spending, so he urges buyers to consider what problem it solves, what evidence proves it works, what happens when it fails, and how the data is governed before making a move.

“I would separate serious AI from weak AI claims with a few simple questions. Where is the system meant to operate, and where is it not meant to operate? What evidence exists beyond the sales deck? What happens in degraded modes? Sensor loss, poor connectivity, bad input, bad weather, fatigue, workload and unusual traffic are not edge details in shipping. They are normal operational concerns,” he claims.

But that is not the only issue for Mihovilović. Another problem with shipping-dedicated AI is the workforce that creates it.

“AI in shipping cannot be built only by software people. It also cannot be evaluated properly by people who have never stood a watch,” he points out. “The key skills are changing. We need people who understand   navigation, bridge resource management, human factors, data governance, cybersecurity and AI assurance.”    As Mihovilović puts it, the valuable person is not only the one who can build an AI model, but also the person who can ask whether the system will still make sense at 3 am, in poor visibility, with a tired officer, a sensor problem, and commercial pressure in the background.

When asked how technology, especially AI, will benefit the future of shipping, Mihovilović says that AI will benefit shipping most when it reduces risk without reducing professional responsibility. Areas such as decision support, early warning, anomaly detection, predictive maintenance, environmental protection, and administrative workload reduction, Mihovilović believes, are where AI can help the most.   He does not, however, support shipping’s idea of using AI for autonomous vessels as its “one single destination”.

“Coastal, inland, port and deep-sea operations are very different. A system operating on a constrained route with nearby support is not automatically evidence of crewless deep-sea operations. The future I see is human-in-the-loop for a long time. That is not a weakness. It is a realistic transition path.”

When a possible oversaturation of the shipping AI market was discussed, he states that, in marketing language, that is the case. However, in mature, evidence-backed, safety-governed AI for shipping, he still sees the market as very young, and that those solutions currently revolve around voyage optimisation, route-risk assessment, bridge situational awareness, machinery monitoring, document processing, and shore-side decision support.

In the years ahead, Mihovilović claims the most important applications will sit at the intersection of safety, operations and assurance.   “That includes bridge decision support, collision-risk awareness, machinery anomaly detection, fatigue-aware safety support, cyber-risk monitoring and tools that help companies learn from near-misses without creating a blame culture,” he reveals to Maritime CEO.   

And ultimately, according to Mihovilović, the long-term winners of the shipping AI race will not be the systems with the loudest AI branding. They will be systems that can demonstrate safety benefits, define their limits, support human skill, and withstand scrutiny from crews, class, insurers, and regulators.