How AI can help to optimise supply chains under pressure
The flow and ebb of international supply chains is far from new, with the grand medieval churches of the Cotswolds partly paid for by wool exports to continental Europe, centuries before the area became famous for hosting US vice-presidents.
Levels of imports and exports rose worldwide during the 19th century and into the 20th during the first wave of globalisation, until two world wars and a depression dragged levels back to those of a century earlier.
Post-war, globalisation’s second wave saw a tripling in merchandise trade – broadly, the total value of imports and exports of goods – from 17% of global gross domestic product (GDP) in 1962 to 51.5% in 2008, according to World Bank data. This was the year after Apple launched the iPhone, an exemplar of globalised manufacturing with the company currently listing suppliers in 29 countries on three continents.
But 2008 was the year of the global financial crisis which saw the collapse of banks including Lehman Brothers. Merchandise trade fell sharply in 2009 and has zigzagged sideways since, staying below 50% of global GDP.
International trade in goods has been hit by one thing after another: the eruption of Iceland’s Eyjafjallajoekull volcano in 2010, the UK’s vote to leave the EU in 2016, the Covid pandemic from 2020, the Suez canal blockage in 2021 and Russia’s invasion of Ukraine in 2022. US president Donald Trump’s rapidly changing tariffs are just the latest problem.
2010’s Eyjafjallajoekull eruption provided an early example, as its dust plume grounded aeroplanes used by, among others, Kenyan flower-growers to reach European markets. A decade later, shortages of everything from personal production equipment to toilet paper during the Covid-19 pandemic highlighted how supply chains could snap.
Globalisation had “probably gone a step too far”, says Emile Naus, a UK-based partner of Amsterdam-headquartered consultancy BearingPoint and previously head of logistics strategy for retailer Marks and Spencer. It has meant companies relying on suppliers on the other side of the world to deliver goods on time with little margin for error.
“If anything goes wrong, the implications are big. We have built supply chains that are quite fragile,” adds Naus, who knows of a food manufacturer that runs two back-up production sites, which add costs compared to operating solely from its main site, because it worries that a shutdown at that main site could put it out of business.
As well as specific events, the shift of manufacturing jobs to China and other lower-cost countries contributed to previously industrialised countries raising barriers to international trade after decades of lowering them.
“It’s very easy to point to the US with its tariffs, but it is only one example,” Naus says. “You could argue that the Brexit changes in the UK and trade restrictions in Asia are part of the same thing.” Climate change is creating further problems by shifting where some crops can be grown and disrupting shipping through droughts.
Richard Howells, vice-president for solution management at German enterprise resource planning software provider SAP, agrees: “Tariffs, shifting trade policies and regional conflicts are all different forms of disruption. Tariffs are the latest one, but it seems to be never-ending.”
He says that labour shortages and requirements to improve sustainability in some countries, including growing requirements to publish data on greenhouse gas emissions by suppliers and customers, add to the challenges.
Speeding the supply chain
Companies have responded by doing things faster. “As the clock speed of supply chain has got quicker, we have moved from making monthly planning decisions to daily planning decisions and hourly in some cases,” says Howells, who has worked in supply chains for more than 35 years.
He says that one SAP customer, a skylight manufacturer, initially thought people would have little interest in home improvements at the start of the pandemic and so planned accordingly. But the opposite was true, with increased demand from people aiming to brighten up their new home offices, causing the company to adjust planning and manufacturing almost daily.
He adds that social media and influencers provide an ongoing reason to have the ability to make changes quickly: “A tweet from a celebrity might trigger that surge in demand.”
Howells says that software suppliers including SAP have supported this acceleration over several years by improving integration between software applications, moving to cloud computing and increasing the use of artificial intelligence (AI). He hosts a podcast on the future of supply chains: “We don’t get through an episode without mentioning AI. I should ring a bell every time someone mentions it. It is a game-changer for supply chain and for businesses in general.”
As well as increasing the pace of forecasting, companies are using AI-driven increased analytical capabilities to widen its scope. For example, rather than making a single forecast some now undertake scenario planning, such as to model the impacts of planned American tariff changes then making decisions to mitigate these.
Howells says this can lead to companies changing suppliers, moving inventory and buying things earlier, although they need the ability to reconsider if or when things change again: “It’s a continuously evolving environment and you have to have business systems and technology in place that can help you predict, react and respond to those things.”
Rohit Tripathi, vice-president for industry at Finland-based supply chain planning platform Relex, says an importer could react to American tariff changes by shifting to supply the US from a country with higher costs but lower tariffs, adding the tariffs to consumer prices or taking them out of profits.
But as well as reacting to specific events, such disruptions are causing customers to reshape supply chains permanently. Rather than “offshoring” to wherever in the world is cheapest, many are exploring “friendshoring”, a preference for politically friendly countries, and regionalisation or “nearshoring”, where companies shorten supply chains. The latter can shorten lead times making a company more responsive to changes in demand as well as reduce transport costs, lower risks of disruption from transport problems such as a blocked canal and cut greenhouse gas emissions.
“Businesses in general are shifting from a just-in-time model to a just-in-case model,” says Tripathi. “What that means is that extreme optimisation of inventory and supply chain is taking a back seat to building in the capability to deal with just-in-case things.”
Technology can help support such regionalisation. German building materials maker Knauf faces competition both on cost and availability as if it turns down an order a rival is likely to accept it and its customers demand reliability, as delays can stop construction work taking place.
“They need to be able to answer the question, ‘Can I have some product?’ and reliably say yes with a date better than their competition,” says Simon Bowes, European corporate vice-president for manufacturing industry strategy at US-headquartered supply chain software provider Blue Yonder.
It also must cope with relatively high costs of transport for some products, meaning it does not make economic sense for it to fulfil orders by moving them between continents. Knauf is working with Blue Yonder to automate 80% of orders by 2032 and manage its supply chain more precisely.
Good supply chain technology also allows companies to launch seasonal or short-term products. Bowes says that it takes at minimum three cycles of data to confirm a trend is not just a blip, so if they plan each month, “that makes it three months before you detect a trend and you’ve missed the summer”.
Dutch brewer Heineken wanted the ability to launch new short-term products and respond to changes in demand without increasing waste through written-off or discounted perishable stock. Bowes says it uses the Blue Yonder’s machine learning capabilities to analyse more data more often than would be realistic otherwise while assessing the changing significance of weather, pricing and competitor promotions for each product.
AI: generative and agentic
Machine learning’s ability to detect trends means that many companies already use this kind of AI in planning. Blue Yonder also sees roles for generative AI, including letting users ask questions and get replies in normal language as well as summarising large amounts of data.
“A warehouse manager coming back after the weekend typically spend the first day catching up with what has happened. We see the role of generative AI and particularly agents to summarise all that information for you quickly,” says Bowes, adding this could help people coming into supply chain work who lack experience of raw data.
AI could also propose solutions to problems, such as listing alternative carriers to one it spots is underperforming. As well as Blue Yonder both Relex and SAP are pursuing this agentic AI approach, with Girteka Group, a Lithuanian-based trucking company, using SAP’s Joule AI copilot for route optimisation to save money and cut greenhouse gas emissions.
However, Bowes believes AI is unlikely to automate supply chain work completely. “We need humans that understand what the AI is doing for them as a means of helping them short-cut some processes, so that they can make their decisions themselves,” he says. “So much of what people do in the supply chain process today is still drudgery.”
BearingPoint’s Emile Naus is cautious about using generative AI in supply chain data analysis. “Put in the right place, it is super-useful,” he says. “Where people talk about it in a supply chain sense, quite often it is misplaced. It’s almost like we have just invented a new hammer and every problem looks like a nail.”
The work requires the use of optimisation and statistical understanding where generative AI large language models “are not particularly strong”, he says, adding that classical machine learning is better placed to do this. “It is a much harder analytical, mathematical approach.”
More broadly, Naus says that companies need to consider data and culture as well as technology when working to improve their supply chains. On data, he says it is one thing to gather good quality data from your own organisation, but quite another to extract it from suppliers – and they may not even want to identify their own suppliers.
On culture, techniques such as sequential decision analysis which replace single forecasts with assigned probabilities for possible outcomes can be very powerful, but difficult for some people to accept.
“It requires you to have a different mindset,” says Naus. “Lots of companies have forecasts for the next couple of years. This is almost saying, ‘I don’t know what the right forecast is’, which is quite counter-intuitive but is much more realistic.”
