In retail and in manufacturing, artificial intelligence (AI) is being put at the heart of back-office operations to flex stock up and down based on demand and to manage supply chain disruption. But many supply chain decision-makers do not see AI as their main priority.
Boston Consulting Group’s inaugural report on the state of supply chain planning, published in February this year, found that volatility and geopolitical disruption dominate the external planning challenges organisations face. In the BCG Annual state of supply chain planning survey 2025, 70% of respondents in a poll of 180 supply chain decision-makers said that demand volatility is their biggest external planning challenge. Global trade and geopolitics is the main external planning challenge for 64% of the people surveyed, while 62% said their biggest planning challenge is supply volatility.
By a long shot, their biggest internal challenge is forecast inaccuracy and misalignment (78%). End-to-end visibility, which is a significant factor for supply chain decision-makers, is a major concern for 38% of the people polled, while AI is only regarded as a top challenge by 14% of respondents. Yet AI is getting a lot of attention, especially in areas such as forecasting.
Forrester vice-president and principal analyst George Lawrie says machine learning is being used in supply chain forecasting to incorporate multiple variables – such as weather, interest rates and housing stats – in demand forecasting models. He says that while traditional methods relied on time-series data, AI enables the inclusion of additional signals for more accurate predictions. In Lawrie’s experience, AI can also assist in sales and operations planning (SNOP) as it can be used to gather and harmonise data from various stakeholders, such as sales, transportation and finance teams.
Managing contracts is another task well-suited to AI. Lawrie says that generative AI (GenAI) can be used for creating and distributing requests for information (RFIs). “It can generate the RFI. This is very tedious stuff; it can distribute the RFI to carriers or suppliers and AI can also help you to evaluate the responses.”
In Lawrie’s experience, organisations use agentic AI to run through the responses received from suppliers from the RFI, by prioritising and evaluating them.
Managing suppliers with AI’s help
RWS Global, which runs and manages live events for its clients, has contracts with thousands of performers, and has been using new AI features built into Box to automate workflows such as the contract approval workflow, saving many hours in terms of manual processing.
When Computer Weekly interviewed Jake McCoy, chief operating officer at RWS Global, in February, he said the automation of the contract approval workflow saves many hours in terms of manual processing: “The end user types in the information that needs to go into a contract via Box Doc Gen, which is then sent over to legal [department] for approval.”
Once approved, the contract is sent out automatically and signed using Box Sign. The signed contract is then uploaded to the cloud. The end-to-end workflow has meant that RWS Global’s contract processing time has been reduced from 20 minutes to under two minutes per contract, reducing what once took more than 8.5 workdays for 200 hires to just five hours.
Using AI with digital twin for troubleshooting
One company that has been using AI in its supply chain is consumer packaged goods giant Unilever, which has been working with Accenture to deploy AI and digital twins across its global manufacturing network. Unilever says it is able to predict 95% of process flow restrictions in deodorant stick manufacturing, delivering a 20% reduction in waste and a 10% uplift in capacity.
Discussing how the AI works with digital twins, Vicky Cuthbert, chief product supply chain officer of personal care at Unilever, says: “At our Raeford site in North Carolina, the digital twin sits on top of existing production systems, using AI to monitor what’s happening in real time. It tracks key process parameters, such as temperature, flow and pressure, to quickly flag any abnormalities that could affect quality or slow down production. As it can spot issues early, teams can identify root causes much faster, reducing time spent troubleshooting.”
If an abnormality is detected, the system proactively alerts line teams and ensures a consistent escalation process. Cuthbert says the digital twin provides predictive insights beyond fixed rule-based thresholds. “By analysing multiple factors at once, the digital twin can detect more complex and hidden patterns that wouldn’t normally be visible, helping predict issues before they occur. It also gives teams better visibility across production by tracking batch-level performance, including batch size and flow, helping to optimise throughput,” she says.
From an operational standpoint, the digital twin integrates with existing factory systems, enhancing current automation rather than replacing it. Unilever’s plan is to move towards a more closed-loop system, where AI agents can increasingly recommend decisions autonomously, with human oversight.
AI for more effective retail planning
On the retail side, last year, Debenhams Group – home to brands including PrettyLittleThing, Boohoo, BoohooMAN and Karen Millen – announced it has been working with Peak, the AI platform from UiPath, to roll out intelligent automation, which is providing the retailer with real-time, automated pricing for thousands of products.
According to Debenhams Group, the AI system helps it to respond more effectively to changing demand, seasonal trends and inventory levels. The system is being used to protect margins, minimise excess stock and deliver more competitive pricing to customers.
Debenhams CEO Dan Finley sees AI as a technology that is transforming how the retailer manages stock. He says: “We’re embracing AI to make smarter, faster decisions that simplify our operations and enhance the customer experience. This technology will transform how we manage stock and pricing – especially during the busy festive season – and help us to continue to deliver great value and service across all our brands.”
Debenhams also expanded its partnership with Amazon Web Services (AWS), adding in the cloud provider’s AI tools. It has used AWS’s serverless cloud technology as the platform on which it has developed a marketplace model to facilitate the faster onboarding of third-party sellers. Debenhams says the marketplace model also provides a broader product selection and helps to simplify the purchasing journey for its customers
Among the uses of AI at Debenhams is AI-generated content powered by Amazon Bedrock, which automates product descriptions and translations across tens of thousands of products. According to AWS, the technology also translates these descriptions into six languages automatically, which means products are ready to sell in different countries quicker.
AI is finding a place in the supply chain
What these examples show is that across retail, manufacturing and other sectors, AI is being used in areas such as onboarding suppliers and helping to streamline supply chain operations. Undoubtedly, as businesses become more confident in using agentic AI, there will be plenty of role to deploy such systems to manage supply chains more effectively.
Yet while there is a lot of interest in AI, according to BCG, fully autonomous planning remains an aspiration. “Most value today comes from foundational applications – improving forecasting, exception management, data interpretation and workflow automation – rather than from lights-out planning,” the authors of the Why AI alone isn’t enough report noted.
According to BCG, organisations that attempt to leapfrog through the process of planning maturity by means of AI alone tend to struggle, while those that layer AI deliberately onto stable planning foundations see more durable gains. In other words, like other areas of business operations, AI will not fix a broken process. But as Forrester Lawrie notes, it can streamline sales and operations and GenAI can be applied to speed up laborious processes such as preparing and managing RFIs.
