Marine

Agentic AI in marine

Agentic AI does not wait for instructions. It acts. That is both the promise and the problem for a marine board deciding whether to let a system make decisions without a human in the loop.

By Zakir Hoosen, Director, Fuzzelogic Solutions. Board-level guidance in plain English.

The term everyone is using right now is agentic AI. It means systems that do not just answer questions. They take actions. They plan. They coordinate. They make decisions and execute them without waiting for a human to approve each step.

In an office, that might mean an AI system that books meetings, sends emails, and updates records without human intervention. In marine, it might mean a system that adjusts a vessel's route based on weather data, cargo requirements, and port schedules, all on its own.

That raises a question every marine board should be asking: when does autonomy become risk?

What agentic AI means in practice

Most AI systems in marine today are advisory. They analyse data and present options. A human decides. Agentic AI changes that model. The system decides and acts.

The difference matters. An advisory system that gets it wrong costs a human time to correct. An agentic system that gets it wrong costs money, cargo, or worse.

In the Isle of Man marine sector, most companies are not yet using agentic systems. They are using AI for analysis, reporting, and prediction. That is appropriate. But the technology is moving quickly, and boards should understand what is coming, even if they are not adopting it yet.

The Gartner prediction

Forty percent. That is not a small number. It means nearly half the agentic projects that start will not finish. The reasons are predictable. The governance was not in place. The data was not ready. The cost was underestimated. The risk was not understood.

In marine, the consequences of a failed agentic project are higher than in an office. A cancelled project in a shipping company might mean a vessel operating suboptimally for months while the company figures out what to do next. That costs fuel, time, and reputation.

Where agentic AI fits in marine

There are areas where agentic AI could add genuine value in marine, and areas where it should not go near a vessel.

Potential areas include:

1. Automated compliance reporting to the Isle of Man Ship Registry and classification societies.
2. Real-time voyage optimisation based on weather, traffic, and port conditions.
3. Predictive maintenance scheduling across a fleet.
4. Automated documentation for cargo, customs, and port authority requirements.

Each of those areas has a clear boundary. The action is contained. The consequences of error are manageable. The human can review and override.

Areas where agentic AI should be approached with extreme caution include:

1. Autonomous navigation or collision avoidance.
2. Crew management decisions.
3. Safety-critical systems on board.
4. Any system where the consequence of error is loss of life.

The Isle of Man government has not yet issued specific guidance on agentic AI in marine. The JFSC issued AI governance guidance in July 2026 for financial services. Marine boards should apply the same principles: someone must own the decision, the system must be explainable, and there must be an off switch.

The governance challenge

Agentic AI breaks most existing governance frameworks. Those frameworks assume a human makes the decision and the system supports it. When the system makes the decision, the governance framework must account for that.

A marine board governing agentic AI needs to answer three questions that do not arise with advisory AI.

First, what is the system authorised to do? Not what it can do technically, but what the board has approved it to do. The boundary must be explicit and documented.

Second, what happens when the system acts outside its authority? Every agentic system needs a containment mechanism. If it cannot act within its authority, it must stop and alert a human.

Third, who is accountable for the system's actions? Not the vendor. Not the IT team. A named person on the board or reporting to the board who answers for what the system does.

If your organisation is in that 21%, agentic AI is not for you yet. Get governance in place first.

The Isle of Man marine opportunity

The Island's marine sector is built on trust. Vessel owners register here because they trust the registry, the regulation, and the legal framework. That trust extends to how the Island approaches new technology.

A marine company on the Island that adopts agentic AI with proper governance, proper testing, and proper accountability sets an example. That example becomes a selling point for the registry and for the Island's broader marine proposition.

This is not opinion. It is commercial reality. The global marine market is watching how jurisdictions handle autonomous systems. The Island has a chance to lead.

What a board should do today

If your board is being asked about agentic AI, the answer is not yes and the answer is not no. The answer is not yet, until governance is in place. That is the responsible position.

Start with the assessment. Two to four weeks, fixed price, and you own the verdict and the roadmap whether or not we build any of it. When you are ready to talk AI, call Fuzzelogic Solutions and ask for Zak. www.FuzzelogicSolutions.com | info@FuzzelogicSolutions.com | +44 (0)1624 618950