From risk to resilience: human + AI collaboration for safer fleet operations

Maritime operations are entering a period of heightened complexity. Sea lanes are more congested, geopolitical tensions are influencing trade routes and GNSS interference has become a real-world navigational threat. Against this backdrop, a joint Digital Ship and Orca AI webinar – From Risk to Resilience: Human + AI Collaboration for Safer Fleet Operations – examined the practical steps fleets can take to strengthen navigational resilience.

Moderated by Nick Chubb, the discussion featured Panagiotis Drosos, Managing Director of Capital Ship Management; Alexandros Politis-Kalenteris, Deputy Chief Operating Officer of TMS Cardiff Gas; and Dor Raviv, CTO and Co-founder of Orca AI. While AI-driven decision support tools were central to the discussion, the message was consistent: resilience depends on culture, communication and measurable performance.

Increasingly demanding operating environment

For TMS Cardiff Gas, the scale of change is significant. The LNG fleet has grown from roughly 400 vessels a decade ago to more than 700 today, with further expansion expected. More ships mean denser traffic, while shorter contracts and generational turnover mean less cumulative experience per officer.

“This is where the majority of serious incidents may occur,” Politis-Kalenteris said, referring to navigation. Near-misses are inevitable; the priority is identifying, analysing and sharing them quickly across the fleet.

For Capital Ship Management, Drosos linked operational excellence to a clear objective: zero incidents. While absolute elimination of near-misses is unrealistic, systematic data collection and structured training can steadily reduce risk.

Connectivity enables data-driven safety

Improved VSAT infrastructure and LEO satellite services such as Starlink – now widely deployed across the global fleet – have made real-time ship-shore data exchange routine. That connectivity underpins AI-enabled situational awareness, allowing operators to move from isolated ship-level decisions to fleet-wide pattern recognition.

Supporting judgement, reducing cognitive load

Raviv emphasised that AI is not designed to replace seafarers but to support them. Modern bridges combine multiple sensor inputs, traffic scenarios and regulatory demands. AI systems can aggregate and filter that data, highlight high-risk targets and identify emerging close-encounter patterns.

The objective is enhanced situational awareness and earlier decision-making – reducing cognitive strain without removing human authority.

Culture before code

Both shipowner execs stressed that technology alone does not improve safety. Culture is decisive.

Initial resistance to new tools, particularly concerns about monitoring, was openly acknowledged. At TMS Cardiff Gas, acceptance rose from roughly 50 percent to more than 90 percent following sustained communication, onboard visits and a clear no-blame approach. When near-misses were identified, the response was mentoring and additional training – not punishment.

Drosos reinforced that culture defines behaviour “when no one is watching”. Transparent data-sharing has replaced the outdated notion that what happens onboard stays onboard.

Managing alarm fatigue

Audience questions addressed the risk of false alerts triggered by sea spray, glare or exhaust. The panel acknowledged that poorly implemented systems can create fatigue. Proper calibration, structured rollout and crew training are essential to ensure systems filter noise and elevate meaningful risk indicators rather than adding to workload.

Measuring what matters

Rather than overwhelming users with raw data, Raviv highlighted two core KPIs: reduction in close-encounter events and increased minimum passing distance.

Across a fleet of more than 1,000 vessels using Orca AI, aggregated data indicates an approximate 50 percent reduction in close-encounter events over time. While causality is difficult to prove conclusively, operators reported clear directional improvements following deployment.

Both shipowners acknowledged that quantifying return on safety investment is complex. However, trend-based performance monitoring provides tangible evidence of progress.

Data governance and knowledge sharing

The panel clarified that operational data generated onboard remains the property of the shipowner. Technology providers analyse data to generate insights but ownership and responsibility sit with the operator.

Digital platforms also accelerate fleet-wide learning. Near-misses that might once have remained onboard are now visible ashore in near real time, enabling faster intervention and more targeted training.

Insurance and regulation

Insurers are beginning to recognise the value of greater transparency and risk reduction, with some P&I clubs supporting adoption of situational awareness and other tools. Over time, parallels may emerge with technologies such as Radar and ECDIS, which evolved from optional aids to industry standards.

Raviv also pointed to Japan’s proactive regulatory approach, where collaboration between government, class and industry is accelerating qualification of AI-based perception and collision-avoidance systems for autonomous-ready vessels.

Key takeaways

As the session concluded, three themes stood out:

• Culture first: leadership must visibly support transparency and continuous improvement.
• Communicate clearly: crews must understand that AI augments judgment rather than replacing it.
• Measure performance: simple KPIs provide a foundation for sustained improvement.

In summary, in an environment shaped by congestion, digital interference and rising operational complexity, navigational resilience is no longer optional. The path from risk to resilience lies in combining human expertise, AI decision support and analytics insights and a safety culture committed to continuous learning.