Maritime artificial intelligence is moving into a more demanding stage as shipowners, charterers and operators look beyond pilot projects and focus increasingly on measurable commercial returns.

The scale of AI activity across shipping has expanded rapidly. Lloyd’s Register said 420 organisations were active in maritime AI development over the past year, compared with 276 a year earlier.

However, wider deployment continues to lag behind experimentation. Research by Thetius and Marcura found that 81% of maritime companies surveyed were experimenting with AI or running pilot projects, while only 11% had formal policies and governance structures in place to support scaling.

The same research found that 33% had moved beyond pilots into operational use, optimisation or broader deployment.

For technology suppliers, the commercial requirement is becoming clearer. Shipowners and operators increasingly want defined baselines, transparent benefit calculations and credible payback periods before committing to wider implementation.

Fuel optimisation provides a measurable case

Fuel and voyage optimisation remain among the most straightforward areas in which technology performance can be assessed against operating costs.

Cargill began working with ZeroNorth on voyage optimisation in 2020. According to Cargill, all of its time-chartered voyages used ZeroNorth technology in 2025.

The company has said the technology helped reduce fuel consumption and costs while supporting schedule performance.

The relationship expanded beyond an initial optimisation project. In 2023, ZeroNorth announced a three-year agreement under which it became Cargill’s primary software provider for vessel and voyage optimisation.

A further development followed on 4 August 2026, when ZeroNorth introduced Propel, an agentic AI system for maritime operations.

Cargill, Ultrabulk and CMB.TECH were named as launch partners. According to ZeroNorth, the first application is a voyage optimisation agent, with the system intended to expand across voyage, vessel and fuel operations.

The latest development adds another AI application to an area where fuel consumption, voyage performance and operating costs can already be measured.

Navigation AI moves into wider fleet deployment

Navigation technology is also being assessed against operational and financial results.

Orca AI has reported that Seaspan vessels using its navigation platform achieved fuel cost savings of about $100,000 per vessel annually.

According to Orca AI, reductions in unnecessary manoeuvres and course changes contributed to the reported savings.

The company also reported a 37% reduction in close encounters and a 35% increase in minimum passing distance in its Seaspan case study.

These figures were published by Orca AI and should not be treated as independently verified industry benchmarks.

The deployment nevertheless expanded. In March 2025, Orca AI announced that another 100 Seaspan vessels would be equipped with its navigation technology.

The case provides an example of an AI system moving from an initial deployment into broader fleet use, while also showing the importance of distinguishing supplier-reported performance figures from independently verified results.

Geneva Dry puts commercial value under scrutiny

The commercial value of maritime technology was a major subject at Geneva Dry, held on 28–29 April 2026.

Daniel Weiss, general manager of shipping strategy at Vale, said some technology products are developed before a clearly defined operational problem has been established.

Vale described an approach in which operational problems are identified first and technology is then selected to address them.

Weiss also said Vale had deployed a machine-learning ETA prediction tool that achieved 70% higher accuracy than the company’s previous methods.

The example shows that measurable AI outcomes in shipping are not limited to bunker savings. Accuracy improvements in operational information can also be tracked against previous methods.

Technology integration was another issue raised during the Geneva Dry discussions.

Christoffer Svard, chief commercial officer of Sea, said maritime workflows contain too many elements for a single technology provider to cover every function and highlighted interoperability between systems as an important requirement.

Sanjay Kapoor, chief executive of GeoServe, also identified connectivity between systems as an important part of maritime digitalisation.

Data quality remains a barrier to scaling

The ability to expand AI across fleets continues to depend heavily on the quality and structure of maritime data.

Lloyd’s Register placed shipping’s overall digital maturity at 2.1 out of four in its Digital Maturity Index, while data standardisation scored 2.45.

The organisation said inconsistent data practices remain a barrier to obtaining greater value from digital technology.

Operational information in shipping is still frequently entered manually or stored across separate systems. This can reduce the quality and consistency of information available for digital analysis.

Lloyd’s Register has also warned that poor underlying data can limit the effectiveness of advanced analytics and AI systems.

The issue becomes particularly important when AI tools are used in voyage planning, vessel performance analysis or other operational processes that depend on reliable input data.

Data quality was also raised during Geneva Dry, where participants discussed the absence of consistent data-quality management in parts of the shipping industry.

For operators seeking to verify financial returns from AI, poor data can also make it more difficult to establish reliable performance baselines.

Demurrage and bunker savings widen the ROI discussion

Measurable returns from maritime digitalisation extend beyond fuel and navigation.

Klaveness Digital has cited a customer that reduced demurrage from high double-digit levels to single digits after digitalising its workflows over a two-year period.

The two-year timeframe is significant because it shows that measurable results from digital implementation may develop over an extended operating period.

Managing director Ingrid Kylstad has also argued that technology purchasing discussions can become too focused on individual cost lines rather than wider operational outcomes.

Bunker operations provide another example where results can be documented directly.

VTS Shipping said it delivered $4.6 million in verified bunker savings to clients in 2025, with recoveries documented operation by operation.

Captain Ali Ihtiyaroglu, managing partner of VTS Shipping, said larger operators were connecting data investment with commercial margins and committing capital accordingly.

These examples demonstrate the different financial measures being used across maritime technology applications. Voyage optimisation can be assessed against fuel consumption, digital workflows against demurrage, and bunker services against documented savings or recoveries.

Governance has not kept pace with AI adoption

Technology performance is only one part of the scaling challenge.

The Thetius and Marcura research found that 66% of maritime professionals surveyed were concerned that excessive reliance on AI could weaken human judgement.

A further 69% were concerned that AI systems could miss important warning signals in areas including contracts and voyage planning.

Those figures sit alongside the finding that only 11% of surveyed companies had formal policies and governance structures for scaling AI.

The gap between experimentation and governance remains one of the clearest signs that maritime AI adoption is developing faster than some organisations’ internal frameworks for managing the technology.

IMO pushes maritime digitalisation forward

The wider regulatory environment is also increasing the role of digital information in shipping.

The International Maritime Organization began developing a global maritime digitalisation strategy in 2025.

In 2026, the IMO Facilitation Committee approved the draft Strategy on Maritime Digitalization in principle and forwarded it for further consideration. Adoption is targeted at the IMO Assembly in 2027.

Shipping companies are already required to manage large volumes of operational and emissions-related information under frameworks including the IMO Data Collection System, EU MRV, EU ETS and FuelEU Maritime.

Digital platforms are increasingly being used to process and manage some of that information.

ZeroNorth, for example, offers emissions analytics covering operational, fuel and voyage data alongside functions related to CII, EU ETS, FuelEU Maritime, EU MRV and IMO DCS.

Autonomous shipping adds another digital framework

The IMO also adopted the non-mandatory International Code of Safety for Maritime Autonomous Surface Ships, known as the MASS Code, in May 2026.

The code took effect on 1 July 2026.

It establishes a goal-based framework covering the safe design, operation and certification of remotely controlled and autonomous commercial ships.

The current MASS Code is voluntary, with the IMO planning a future transition toward a mandatory instrument.

The code does not establish a commercial framework for AI investment, but it forms part of the wider regulatory development surrounding increasingly automated maritime operations.

From pilot projects to documented results

The maritime sector has not lost interest in artificial intelligence.

The increase from 276 to 420 organisations involved in maritime AI development shows that activity continues to expand.

However, the contrast between 81% of companies experimenting with AI and only 11% having formal structures for scaling it shows that adoption remains uneven.

Commercial deployment is increasingly being assessed use case by use case.

Fuel optimisation can be measured against bunker consumption. Navigation systems can be assessed through manoeuvres, deviations and safety data. ETA tools can be compared with previous prediction methods. Digital workflow platforms can be examined against demurrage, while bunker management services can be measured through documented recoveries.

The next phase of maritime AI adoption will therefore depend increasingly on evidence that individual applications can deliver identifiable results under real operating conditions.

For shipowners, operators and charterers, the issue is no longer simply whether AI can be deployed.

The commercial question is whether the technology can produce results that can be measured and justify wider investment.