
As AI use becomes more commonplace in maritime operations, the role of humans in the loop must be clear and actively designed to be empowered, writes Gleb Tsipursky, PhD
A port official receives a summary showing that a vessel’s arrival documentation is complete. The record appears orderly: crew details, cargo declarations and clearance status in one place. An hour later, an agent discovers that the source system updated a certificate after the summary was generated. Does the team know which version went into the decision, who can correct it and which authority must be informed?
This is a hypothetical scenario. Yet the underlying question is becoming more relevant as maritime information moves through more connected systems. Digitalisation can spare people repetitive work. It can also make it harder to see the human judgment behind a seemingly finished record.
In September 2026, the International Maritime Organization (IMO) launched a survey asking maritime stakeholders about their experience implementing Maritime Single Windows. These systems have been mandatory for IMO Member States since January 2024, providing a central electronic channel for information exchanged between ships and public authorities during port calls. That is a substantial change in how shipping paperwork travels. It does not mean that every record is now correct, or that every port agency interprets the same information in the same way.
The gap between transferring information and understanding it deserves more attention in maritime AI discussions.
Faster forms vs better decisions
Imagine a shipping agent using AI to prepare a plain-language summary of a vessel’s arrival documents. The system can highlight omissions, group records and suggest which questions to resolve. These are useful activities because information may have to reach multiple agencies under time pressure.
But an AI-generated summary can also flatten distinctions that matter. One date may record when a certificate was issued, another when a record was entered and another when an update became effective. An assistant might present the most recent date as the only date. It might combine two names that look similar but identify different parties.
Human expertise becomes especially valuable at that boundary. The relevant skill is not manually retyping the information. It is knowing which assertion requires a source check, which difference is consequential and when a discrepancy needs to move to an authorized decision-maker.
That skill can weaken if employees see only polished outputs. A trainee who once compared the underlying documents may now learn to accept an answer without understanding its construction. Experienced staff may be able to catch a subtle inconsistency, while newer colleagues have fewer opportunities to practice.
Treat that as a design problem, not a reason to reject automation.
A handoff that someone can explain
At its March 2026 meeting, the IMO’s Facilitation Committee approved an overarching strategy on maritime digitalisation. The organisation describes priorities including interoperability, data sharing, data governance and resilience to cyber threats and disruption. The strategy is moving through further IMO consideration, with Assembly adoption planned for 2027. This remains an evolving framework, not an already adopted universal technical standard.
Human understanding belongs alongside those technical priorities.
For each important AI-assisted step, a port or shipping organisation can record four practical facts. First, what original source supports the output? Second, when was the relevant information last confirmed? Third, what is the system permitted to prepare or recommend, and what requires an authorized human decision? Fourth, who can pause the process, resolve a conflict and inform other affected parties?
These are proposed operational practices, not IMO requirements.
A useful test is whether a staff member outside the software team can explain one unexpected output. If a summary says a declaration is complete, can the employee identify the underlying entries and the responsible authority? If two records conflict, can that employee explain why the system cannot safely settle the matter by guessing?
Transparency becomes practical when it gives someone the means to act.
Practice the exception
Maritime digitalisation programmes often have compelling demonstrations: one submission, less duplication and a cleaner workflow. Training should also show what happens when the information does not fit.
A short exercise could provide a fictitious port-call packet with a vessel arrival notice, a certificate and an amended entry from another agency. Introduce one conflict in the dates and one missing authorization. Ask participants to review an AI-prepared summary, mark the statements that lack support and decide which issue they can correct and which they must escalate.
Then change the facts. Make the amendment arrive after the initial summary. Give the team conflicting instructions from two fictional agencies. Ask what needs to be logged and who should receive a correction. The scoring should reward accurate escalation and identification of uncertainty, not simply speed.
The point is to make judgment observable. Leaders can measure whether staff trace claims back to records, notice altered information and identify the responsible party. Those are proposed training measures, not evidence that a particular port has achieved them.
There is also an organisational dimension. Junior staff need a safe way to say that a fluent machine-generated response is unsupported. Supervisors need to make it clear that reporting an inconsistency is part of good performance. If everyone is rewarded only for clearing the queue, unresolved questions may be pushed downstream.
Digitalisation is a coordination challenge
This issue extends beyond individual employees. In October 2026, the IMO reported work with Senegal to assess the steps needed for a Maritime Single Window at the Port of Dakar. The mission engaged port administrators, government authorities, shipping lines, agents and terminal operators. That range of participants illustrates how a shared digital process crosses organisational boundaries.
A successful handoff should identify both a data destination and a person or authority responsible for resolving exceptions. Otherwise, an error can travel faster while accountability moves more slowly.
The maritime sector has good reasons to value quicker information exchange. UN Trade and Development’s 2025 Review of Maritime Transport also warns that automation and digitalisation can increase cyber risks, alongside their efficiency potential.
The next useful question for maritime leaders is therefore sharper than how many minutes an AI tool saves. When the information is wrong, incomplete or disputed, can the people operating the system still understand the record, identify the limits of their authority and correct the course?
A digital port should have faster documents and stronger judgment. Both need deliberate design.
——
Photo by Igor Omilaev on Unsplash. Words: Gleb Tsipursky, PhD, behavioral scientist, CEO of Disaster Avoidance Experts, and author of The Psychology of AI Adoption at Work: From Resistance to Results (Georgetown University Press, 2026).
—————–
Like this article? Click here to sign up for our free monthly ‘De-Mystifying Digitalisation and AI in Maritime’ newsletter and get the latest cross-sector expert insights, news, interviews, videos and case studies straight to your inbox.