According to Gartner, Generative AI will power close to 25% of all logistics KPIs by 2028.
AI is also predicted to make supply chains 45% more effective in timely and error-free product delivery (Research and Markets), with half of organizations predicted to be using AI-enabled tools to support their supplier contract negotiations by the year 2027.
While automation and AI are now considered critical drivers of change and transformation, many teams remain at an impasse, lacking the internal skillset, technical maturity, or data available to deploy or scale AI-backed solutions.
Instead, organisations are using 2024 to build use cases and conduct pilots – with some of the more popular use cases being contract risk analysis or operational tasks. This second use case was supported by customer anecdotes shared at the recent Xeneta Customer Advisory Board (XCAB) AI roundtable (May 22nd).
The rest of this article will focus on takeaways captured during that roundtable – and some bonus top tips.
Current AI usage
Those in attendance all identified themselves as interested observers of AI and emerging technologies, but there was minimal AI or GenAI usage today. Those who had engaged with AI-based solutions were focused on operational tasks such as forecasting, reducing the manual lift from procurement teams and supply chain optimizations.
Interestingly, while AI expansion was identified as a business goal, its potential for transportation wasn’t yet a high priority. That said, one participant was actively seeking out partners who support the development of AI in their freight procurement and supply chain management – something which will help advance operations and sustain their position in the market as one of the largest exporters in their region.
When it comes to supply chain procurement optimization, GenAI was seen to take priority. But even then, brand messaging and internal processes came first. As one participant mentioned:
“We’re starting to look more into how we can use that technology with ocean contracts because we do manage so many different commodities and contracts. There is a heavy lift on our teams to do these manual processes.
It’s shocking to me how companies as large as ours operate in Excel spreadsheets, but that is the case.”
Reservations
Reservations for using AI center on security, reliance on third parties, and a lack of maturity around AI use cases that calculate and predict unforeseen circumstances. There’s also a lot of unknowns about what use cases are possible and the expected ROI, which is resulting in a lack of trust in AI.
As one of the participants noted:
“The moments when you need the model to be the most accurate, it hasn’t learnt enough, so what you have to do is go in and manually adjust the model – that’s the other thing I’m a little bit cautious about when using AI and predictive [modelling] in transportation because there’s no model that’s predicts the Houthi rebels firing at ships in the Red Sea.”
Another shared that being a privately owned company means that while they are being encouraged to ‘find efficiency through technology’, they must adhere to strict confidential policies. This has seen them develop their own internal version of ChatGPT to stop them needing to trust external platforms.
Difference between AI and Generative AI (GenAI)
AI has been around for a long time. But it’s the hype around GenAI that’s increased the focus it gets in the boardroom.
As Fabio Brocca, Chief Product Officer at Xeneta, shared at the roundtable:
“I was looking at the Gartner hype cycle which looks at emerging technologies and where we are in terms of adoption. When you look at GenAI – we are at the peak of the hype. There’s a lot of traction, everyone’s talking about it, but it’s very hard to understand the real use cases and impact that GenAI will have on our businesses”
Brocca also stressed the importance of understanding the difference between AI and GenAI.
In short, AI focuses on performing a specific task or responding to a particular set of inputs in a way that is normally associated with human intelligence. AI can learn from data, draw insights, and make predictions.
GenAI on the other hand is a subset of AI and can create content based on ‘prompts’ or inputs. This is possible thanks to the underlying training datasets and deep learning models that GenAI is based on. Whilst the new content will resemble its training set, its generative nature enables the model to learn relationships and patterns and evolve these into new works of text, imagery, code, video...
Do your homework before committing budget
Understanding what problems AI can solve is an important distinction to make before committing KPIs and budget to AI technology. Do you want your AI to be creative and produce new viewpoints based on synthesized information, or would anomaly detection and predictive analytics be of greater business value?
Once you’ve established what AI can do for you, you need to understand if your organization has the right level of maturity, talent, appetite for change and internal capability to start exploring use cases. If you’re ready for use cases, be sure to assess how any new technology will integrate with your existing tech stack, and the potential impact on existing teams, tools and processes before scaling successful pilots.
And perhaps most importantly, you have to make sure that the outputs of your AI solution are explainable and traceable back to an algorithm that can also be explained.
This starts by evaluating the quality of your data.
Do you have the right data or large enough datasets to achieve full AI effectiveness? Has your data been cleaned and aggregated in a way that will appropriately feed AI-backed insights? Or are you currently concerned about facing a ‘garbage in, garbage out’ scenario?
As an attendee noted:
“As far as AI goes, I think there’s rapid adoption without necessarily the same level of understanding of what people are adopting. When you see lawyers that are quoting case laws that don’t exist because an AI model told them so… ‘I always ask the question, what was your model trained on?’ That becomes the key factor.”
Is AI always the answer?
While some shippers have been using AI to successfully support predictive ETAs, the ability to accurately adjust AI models during times of major industry events has caused these early adopters to question if AI can truly keep up with a constantly evolving shipping landscape.
For one of our attendees, model adjustments during the pandemic were unfeasible. Not only were there too many market variants, but there wasn’t enough historical data to know what to adjust their models to. And back to an earlier point, “there’s no model that’s predicts the Houthi rebels firing at ships in the Red Sea.”
In response to these concerns, Brocca agrees that there is currently no AI model capable of predicting black swan events, because by their nature, they are unpredictable. But he does believe
“that there is a point where potentially the algorithm will have seen enough data through enough disruptions to be able to quickly adjust to the new scenario and still provide valuable guidance on where the market is going."
Rather than waiting for AI to figure out how to predict the future, Brocca suggests that more attention should be placed on how fast your organization can react to these events when they do happen. For Brocca,
“it’s about making the assumptions that you use in your model visible and explainable. And when one assumption changes because you have a black swan event, adjust your assumptions or show why they are now invalid. And maybe in these situations, the adjustment is that we don’t rely on AI anymore, but we leverage the knowledge of our market experts and Xeneta community to navigate the uncertainty of an unpredictable event.”
Future aspirations
There is no denying the potential of AI for transforming global supply chains into customer-centric and proactive, but when 40% of innovations fail (Gartner), there are steps organisations need to take to support any digital transformation ambitions. And in the case of AI, slow and steady might win the race.
For the roundtable attendees, future business aspirations currently sit at strategy/management levels, with the biggest hope being that AI will soon support the following:
- Predicting freight rates in relationship to global events and lane-level specifics,
- Predicting rate drivers (current events),
- Demand planning,
- Production estimations,
- Predictive analytics,
- Index adjustments,
- Digital twins to learn from and test potential scenarios on historical data,
- Data maintenance.
Our customers are also curious to see how carriers and Freight Forwarders will make use of this technology. In particular, how carriers will effect change in contracts and bids process, considering existing challenges around current bid cycles. As one participant noted:
“The value of being able to predict the market is before I sign my contract this year. Am I signing a good contract? Not based just on what everyone else is signing right now, but what future rates are going to look like.
“I question how I would lean on a rate prediction a year out. That’s where the real value to me would be, is this contract rate going to look great for the full term, or do I need to consider adjusting my contract terms?”
More than anything, Xeneta customers want AI-based solutions to give them a better grasp across different trades and different commodities, what are the current drivers in relation to rate challenges and is there a way to help shippers identify not only what the drivers are, but which are the biggest drivers.
Dirty data, wonky information
The danger with the hype around AI in shipping is that “almost everybody can create a black box ML model that is going to spit out a number and call it a pricing prediction.” (Brocca)
But what we’ve learned talking to customers is that context matters, and rate predictions are useless when they can’t be explained. From a customer perspective, you must be able to give reasons as to why you believe the market is going up or down, and what factors are contributing to market movement.
Xeneta will continue to prioritize real-time, quality freight intelligence that’s trusted by the world’s biggest buyers & sellers of ocean & air freight.
Because we understand that without clean data, any model you build is going to give you wonky information.
Next steps
If faster market monitoring and decision-making are important to you, discover the new era of ocean and air freight market intelligence.
Speed. Layered Data. Granularity. Context.
A view of everything impacting your deliverability and bottom line, all in one place. So, you benefit from real-time market highlights and personalized insights to simplify reporting, streamline tender processes and help you identify where to focus when disruption hits.
Discover the next level of freight intelligence today and start modernizing the way you buy and sell freight.
Feature image designed by Freepik.
