You may already use AI to summarise a supplier document, prepare an email or make sense of a technical report. It saves time, the answer reads well and the next task is waiting.
But before that answer becomes an instruction sent to a vessel, who checks that it is correct?
On 30 September, Thetius and Marcura published Earning Trust: AI in Maritime. In a survey of 60 maritime professionals, 63% reported using AI daily, while only 8% described their organisation’s use as mature and governed.
It is a small sample, but it raises a practical question about how teams manage the tools already entering their working day.
Imagine asking AI to shorten a maintenance instruction before forwarding it onboard. The result is easier to read, but an important condition has disappeared. The person receiving it may never see the original document or know that AI helped prepare the text.
The level of checking should reflect the consequences of an error. Polishing a routine email and interpreting a technical procedure need different levels of review.
Where an answer could affect equipment, safety or an operational decision, someone with the relevant knowledge needs to check it against the original source.
A useful starting point is to make responsibility clear. The person using the answer should know who has reviewed it and whether it applies to the vessel and equipment concerned.
The same attention is needed before information enters an AI tool. A support request may contain network details, a crew document may include personal information and a supplier report may be confidential.
Teams need practical guidance on which tools are approved and what information they can share.
At Sea IT, years of working with vessel IT environments have taught us the importance of understanding how people actually use technology. Working closely with shipping companies and onboard teams helps us connect technical controls with everyday routines.
That experience is relevant as AI becomes part of maritime work. Approved tools, appropriate access and clear responsibility give people a foundation for using new capabilities confidently.
Reviewing real examples with the team can also reveal where guidance needs to be clearer. Can someone recognise when an answer needs further checking? Do they know whom to ask? Does the process still work when the usual contact is unavailable?
Define
Make approved tools, permitted information and responsibility clear.
Verify
Check important outputs against original sources and vessel circumstances.
Improve
Use mistakes and feedback to strengthen guidance and everyday routines.
Before forwarding the next AI assisted answer, ask yourself whether the person receiving it can rely on the checks behind it.
