WHILE general AI engines have become powerful tools across the market, they are not adequate for maritime’s complicated workflows, believes Russ Hubbard, chief commercial officer of Veson Nautical. In this podcast, he says the huge data sets used to train generalised tools do not include shipping-specific information, which makes data — both company-sourced and shared information — “foundational” to making AI truly effective in maritime.

“You need something that intrinsically understands the industry,” he says, because shipping has its own jargon, processes and workflows that are not addressed by the readily available large language models used in the broader market.

His view is based on Veson’s experience both from building AI-powered solutions for clients and from using AI in the team’s own work. The company recently published a guide on how maritime teams can obtain the most out of AI, which is titled, Maritime’s AI Foundation – Building the Infrastructure for Intelligent Trade. “Our industry is at an inflection point with AI,” Hubbard says.

As an example of a shipping-specific process that is beyond the scope of non-specific AI, he refers in the podcast to preparing demurrage claims, which require AI to have a greater understanding of maritime matters than can be achieved by applying a general AI procedure.

Hubbard also speaks about the extent of unstructured data — such as information contained within the hundreds of emails maritime professionals get everyday — that when structured and organised, can unlock new visibility and opportunities for companies.

Trustworthy data

For that to happen though, Hubbard underscores that data must be trustworthy. He talks of an interaction with the audience during a discussion at the dry bulk conference Geneva Dry in April, during which he asked for a show of hands to indicate if attendees trusted their own data. Their surprising response leads him to suggest that the same attitude probably applies to their trust in the AI analysis of that data, preferring human interpretation instead.

Reflecting on this attitude, he suggests that it is hard “for all of us to admit that there is computing power that can outperform a human in some aspects of the job”, prompting him to consider the effect of AI on future roles. “Once you’ve taken 95% of the tedium out of people’s work... how do you really excel at that remaining 5%?” Hubbard wonders.

The answer may lie in his comments about what AI does not do: “it is not able to do relationships and [cannot] provide judgement based upon personal experience”, he says.

Hubbard suggests four takeaways from this podcast regarding the use of data and its potential to improve competitive advantage. He also reminds listeners to manage their expectations as they embark on AI adoption. “Remember that it’s a journey. It’s very difficult just to go to the highest order of evolution on your first try,” he says.