Tech / AI all very well – but 'with no data layer, nothing will work'

Freight’s AI boom may be gathering pace, but according to cargo.one, the real problem isn’t the sophistication of bots, but the messy, disconnected data foundations on which they are expected to run. 

Speaking to The Loadstar, cargo.one founder and co-CEO Moritz Claussen said the past 12 to 24 months had brought a “wave of companies coming into logistics with AI technology, but with very, very little background on how logistics work”. 

“As a result, you often see very flashy, great demos that work on demo data… but then, as you actually try to put this into operation, very quickly it starts to fail, because the underlying data… is just not connected,” he said. 

“A quoting agent can only be as good as the rate data it has available to quote on.” 

The Berlin-based provider has long been known for airfreight e-booking and rate retrieval, but it announced today it had completed the acquisition of ocean rate platform Cargofive “to unify air and ocean rate data”. 

Part of the update, Mr Claussen said, was rising dissatisfaction with existing forwarder tech stacks – particularly in the TMS space.  

“There’s been a lot of upheaval lately on  what is the value a trade management system provides… how slow and really not very innovative these systems are, and how much are they costing,” he said. 

Cargofive founder and CEO Sebastian Cazajus told The Loadstar customer demand had shifted decisively toward multimodal.

“Every forwarder is doing ocean and air, and nobody wants to have information across different silos or go to multiple sources. So everyone is waiting for this.” 

He added that some customers were already trying to build their own AI portals, but ran into the same wall: “This front end is super nice, this bot is great, but if they don’t have the data layer, nothing will work.” 

According to Mr Claussen, players will benefit from an operational “system of action” that can sit above the trade management system and remove most manual steps in sales and procurement. He explained that this was one of the best use-cases for AI – to be practical rather than performative.  

“What cargo.one does now… is we have all of the rate data in one place,” Mr Claussen said, adding that teams could “quote it automatically in real time”, and then book it with the carrier “directly in one workflow”. 

He urged: “The faster a freight forwarder is able to quote, the more likely it is that they will actually win that quote.” 

But a key challenge is still the industry’s distinction between a price, and a price you can actually execute.  

“It is about accurate rate data… where do you get that rate data from and how accurate is it?” he said. “And the other thing is, do you have a connection to the carrier… through a live digital connection? Because only if you do, will you always be up to date.” 

While Mr Claussen noted that repetitive tasks, such as a “200kg general cargo shipment”, could be executed by AI agents with “minimal oversight”, he explained that “as soon as it goes above a certain size or a certain complexity”, the processes benefited from a “human in the loop”. 

Pricing for cargo.one reflects that split, Mr Claussen told The Loadstar. Standardised use-cases can be sold “on a monthly fee”, while larger deployments follow a process audit and ROI design before moving to “outcome based”. 

“If the job isn’t successful, you’re not paying for it,” he said.