Recent research indicates that 63% of maritime professionals utilize artificial intelligence (AI) in their work on a daily basis. However, only 8% of them recognize their organization as a mature entity with governance in the use of AI.
Costs of Verification and Trust in AI
More than half of the participants (55%) spend at least one hour a week reviewing or correcting outputs generated by AI, and nearly one-fifth of them dedicate five hours or more to this task. Nevertheless, 85% of these individuals have stated that using AI ultimately saves them time.
The so-called “verification tax” report refers to the time and attention spent on reviewing AI-generated work. The main question is how much of this verification is necessary. Trust in AI-generated outputs is still growing; 57% of individuals are positioned in the middle of a five-point scale.
Challenges of Trust and Review
Some reviews are likely to be unavoidable. Seven out of ten say they always review an AI suggestion before acting on a significant business decision, while 80% believe that a specific person should always be responsible for a business decision, no matter how capable the AI is.
The majority of respondents encounter errors while reviewing AI outputs, and fewer than one in ten say their organization's data is fully ready to rely on AI. If experience shows that an AI output is reliable, 47% said they would spend less time reviewing it.
Future Opportunities and Challenges
The report states that the opportunity lies not in eliminating human oversight but in maintaining it appropriately: keeping oversight where the consequences of mistakes are critical and reducing unnecessary reviews. For example, the CP Optimiser tool, which is a review tool for leasing contracts, has been jointly developed by Marcura and Tomini Chartering. This tool analyzes contracts with the company's accumulated knowledge and highlights clauses and risks that may need to be identified by an experienced colleague.
Tomini Chartering's CEO, Thomas Stearn Holm, describes this system as a “collective brain” for the organization and emphasizes that it is a human-to-AI-to-human process: humans input knowledge, AI gathers it and connects it to the relevant clause, and then humans make the final decision.
The conclusion of this report is that companies should measure the complete human and AI process, distinguish necessary verifications from reviews arising from uncertainty, use human corrections to improve future performance, and establish clear ownership and governance.
Read more: Eight U.S. Marines Injured in Iranian Attack on a Ship · New Report Shows 40% of Mariners Feel Unsafe Reporting Safety Issues
