15 Real AI Use Cases Ship Operators Will Pay For In 2026

Routing and speed decisions relied on generic weather routing, static sea trial curves and experience on the bridge, with limited data sharing to shore. Most optimisation was done with simple what-if checks, so only a few route and speed options were tested and results varied by ship, master and trading pattern.

• Performance curves often no longer matched a fouled or retrofitted hull.
• Voyage plans rarely accounted for the full impact of currents, waves and congestion together.
• Racing to port and waiting at anchor was common when port and congestion information was patchy.

Port calls were planned with fragmented data, fixed schedules and a lot of telephone and email traffic between ship, agent, terminal and port. Operators often had a rough idea of congestion and line up, but it was hard to turn that into a precise slow down order.

• Masters tended to steam at or near service speed to avoid being late, even if the berth was not ready.
• Waiting time at anchor could run into many hours, adding fuel burn on the approach and local emissions near the port.
• Line-up changes were shared in different formats and channels, so updating plans across a fleet took time.

Speed and trim decisions depended on sea trial curves, noon reports and experience. Operators ran periodic fuel-saving campaigns, but it was hard to keep the tuning up to date as fouling, loading patterns and routes changed.

• Performance curves were static, often based on a new or clean hull at ideal conditions.
• Trim and draft guidance came from generic tables rather than ship-specific data across many voyages.
• Most “optimisation” happened via manual spreadsheets and offline studies, not on every voyage and every leg.

Most fleets combined OEM recommendations, running hours and local experience. Planned maintenance systems handled the schedule, but they did not know whether a bearing, injector or pump was actually degrading faster or slower than the book expected.

• Critical tasks were mainly based on fixed intervals and hours, not on actual condition.
• Unexpected failures still appeared between services, leading to off-hire, diversions or extra port time.
• Shore teams had limited visibility into how machinery was behaving between noon reports and service visits.

Hull and propeller maintenance was based on rules of thumb, diver photos and bunker curves updated once in a while. Many vessels sailed for months with more fouling than anyone realised, and cleanings were often booked either too late (after big losses) or too early (wasting coating life and cash).

• Speed loss estimates were rough and hard to separate from weather, loading and routing effects.
• Cleaning decisions were driven by calendar intervals, complaints about speed or charter pressure, not hard data.
• It was difficult to compare different antifouling coatings, ports or cleaning methods on a like-for-like basis.

Internal tank, hull and deck inspections depended on people climbing, rafting or using scaffolding in hard-to-reach areas, taking photos and notes as they went. Coverage varied by inspector and ship, and most of the analysis was done manually back in the office.

• Inspectors had to spend long hours in confined spaces or at height, increasing risk and limiting how often checks could be done.
• Defects could be missed or inconsistently graded because different inspectors had different thresholds and methods.
• Photo archives were hard to compare season to season, making it difficult to quantify whether a problem was stable or accelerating.

Navigating in traffic relied on human scanning across multiple displays plus out-the-window visuals. Radar, AIS and ECDIS were powerful tools, but they were loosely connected: each officer mentally fused the information and decided which targets were important and which could be ignored.

• Workload on the bridge spiked in congested waters, bad weather and pilotage, increasing the chance of missed cues.
• Close-quarters risk assessment depended heavily on experience and moment-to-moment concentration.
• Near-misses were hard to reconstruct later because data from different systems was stored and reviewed separately.

Market views were assembled manually from a mix of broker circulars, Excel sheets, email threads and individual experience. Each person maintained their own models and rules of thumb for TCE, positioning and relative value between routes and sizes.

• Answering simple questions (e.g. “Is Pacific or Atlantic better for this Cape in 10 days?”) could take hours of data gathering.
• Insight quality varied widely between desks and people, depending on how disciplined they were about updating models.
• Historical knowledge was often locked in personal spreadsheets and inboxes instead of a shared, searchable system.

Emissions and CII management relied on periodic spreadsheets and static calculators. Data from different systems was stitched together manually, often weeks after a voyage closed, which meant most decisions were backward looking.

• ETS exposure was estimated with rough averages and updated infrequently, making it hard to manage quarterly carbon cost.
• CII scores were checked once or twice a year instead of being treated as a live operational constraint.
• Scenario work for new tech, speed changes or trading patterns meant building fresh spreadsheets each time.

Fuel planning was driven by individual experience and static price sheets. Buyers compared a handful of ports and quotes at a time, mostly in email, and it was hard to see how one stem decision affected total programme cost or emissions.

• Prices were checked for a short list of “usual” ports, so cheaper options with small deviations were often missed.
• Forward price views, ETS cost and quality risk were rarely modelled systematically into stem decisions.
• Procurement history and supplier performance sat in scattered spreadsheets and inboxes instead of one searchable view.

Many stowage plans were created manually or with rule-based tools that left most of the trade-offs to the planner. Experience mattered more than data, and similar ships on similar loops could end up with very different bay plans and handling times.

• Planners juggled stability, stack weights, DG rules and crane productivity mostly in their heads and in spreadsheets.
• Restows, badly positioned reefers and uneven crane workloads were common, especially on complex multi-port rotations.
• Time pressure meant “good enough” plans rather than consistently optimised use of slots and deck space.

Compliance and risk checks were mostly manual. Teams searched sanctions lists, reviewed AIS plots case by case and tried to keep up with advisories from regulators and banks. Many red flags were only spotted after a trade was underway.

• Screening focused on simple list matches, not on behaviour (dark activity, suspicious STS, frequent high-risk ports).
• Ownership and control structures were hard to keep current, especially for fleets with frequent changes.
• Conflict zones, war-risk areas and new regimes (price caps, corridor rules) had to be tracked via separate sources.

Fatigue and wellbeing were tracked mainly through paper or basic digital rest-hour records and occasional surveys. Compliance could look acceptable on paper while real fatigue and stress built up under the surface.

• Rest-hour entries were sometimes completed in batches, making it hard to spot real risk or non-compliance.
• Near-misses and minor incidents were not consistently linked back to fatigue or workload patterns.
• Mental health and wellbeing support was patchy, with crews often unsure where to turn until problems were serious.

Reporting meant copying the same voyage and fleet data into multiple Word and Excel templates. Every regime and stakeholder (regulators, class, charterers, banks, boards) wanted a slightly different format, and small changes took days of manual work.

• High risk of inconsistencies between internal KPIs, MRV/DCS, ETS submissions and charterer or vetting questionnaires.
• Key staff spent a large share of their time compiling and re-checking reports instead of analysing what the numbers meant.
• Historic reports were hard to search or compare, making trend and portfolio analysis slow and error-prone.

Charterparties and trade docs were reviewed line by line in Word or PDF, often under time pressure and across long email chains. Knowledge about “good” and “bad” wording lived in a few senior charterers’ heads and in scattered clause libraries.

• Important changes in riders or small edits to standard clauses could slip through when teams were busy.
• Cross-checking recap, main CP and later amendments for conflicts was slow and sometimes skipped.
• Portfolio view of exposure on demurrage, laytime, off-hire, performance, pollution or sanctions clauses was limited.