Artificial intelligence (AI)-enabled supply chain management has the potential to supercharge demand forecasting, revolutionize end-to-end transparency and boost integrated business planning. According to ASCM’s Research, Innovation and Strategy Committee (RISC) Sensing Subcommittee, AI and machine learning (ML) are among the 10 supply chain trends to watch in 2022. “They are foundational to integrating people, processes and systems in a wide array of operational environments,” the research states. “The technology-driven evolution to industry 5.0 — which involves a more collaborative approach, as well as partnerships between humans and robots — will have significant impact on supply chain functions such as planning, demand management and fulfillment. As machines learn, improved insights will be discovered, leading to significant transformation, advancement and competitive advantage.”
Indeed, the size of the prize is hard to ignore. McKinsey’s Succeeding in the AI Supply Chain Revolution estimates that AI and ML’s ability to support better decision-making through the analysis of massive volumes of data has enabled early adopters to improve logistics costs by 15%, inventory levels by 35% and service levels by 65%.
Five exciting applications in supply chain
AI is about making better decisions faster by combining significantly more inputs and signals than humans can process. AI can adapt in near-real-time to changing conditions and develop new knowledge by processing more data and revealing hidden correlations. This has led to the rise of several game-changing AI and ML applications, with countless more to emerge in the future:
1. Autonomous delivery robots and drones are being used for last-mile delivery, slashing costs, reducing the traffic burden on roads and improving delivery times. These machines can handle navigation, trajectory adjustment, moving obstacle detection and avoidance — all in near-real time, says Desirée Rigonat, PhD., optimization and machine learning consultant at DecisionBrain. “Technological awe aside, autonomous delivery has proven incredibly useful during the pandemic,” she notes.
Further, according to a 2020 Forbes article, contactless delivery for medicines and groceries has been a much sought-after service in the United States, and companies that offer delivery drones and robots have seen their business grow.
2. Warehouse workers of the future will be increasingly equipped with augmented reality tools, such as smart glasses that enable hands-free order picking. This also has enormous potential for improving warehouse efficiency, as illustrated in a recent pilot developed by Ricoh and DHL.
3. AI and ML have the potential to boost multi-dimensional decision-making processes that take into account impacts beyond the walls of an organization, such as environmental, social and governance (ESG) applications. “AI enables a longer-term approach through models and simulations and can take over data-management processes such as data discovery, collection and processing,” says Bertrand Maltaverne, procurement digitalist and senior analyst at Spend Matters. “AI and ML can act on this data, either autonomously or by providing human decision-makers with data-based recommendations.”
4. Digital twins enable supply chain management professionals to test the impact of a change in a zero-risk virtual environment before implementation in the real world. Maltaverne says they can be used to design supply chains, analyze scenarios, build knowledge and optimize operations. Users can create proactive optimizations based on real-time signals — demands, markets and geopolitical — and, when incidents happen, either anticipate or react immediately via contingency plans or ad-hoc recommendations.
5. AI and ML in supply chain also can enable smarter process automation, both upstream (supplier discovery, supplier qualifications and sourcing) and downstream (inventory management, order management and freight optimization). As McKinsey notes, automation of the physical flow of goods is built upon prediction models and correlation analysis to better understand causes and effects.
Getting the most out of AI and ML as predictive tools
Supply chains can maximize their investment AI and ML through careful change management, data quality and availability. “Organizations must have a change-oriented mindset,” Rigonat advises. “In most cases, AI technologies radically change the way people work, so it is fundamental that involved workers and their management are receptive to this change.”
Communication is key: Employees who believe the new tool is there to replace them will not contribute to its adoption and improvement, which will ultimately cause the project to fail. People should be convinced that the initial learning curve is a fair price to pay for the value they and their teams will receive from these technologies.
In addition, many organizations underestimate the time and effort that will be involved in ensuring data quality and availability when transitioning to an AI-based solution. “If the data is imprecise or incomplete, the tool will not be able to produce useful results, following the well-known garbage-in garbage-out principle,” Rigonat warns.
Maltaverne agrees: “The highest cost for an AI initiative is not the AI itself – in fact, there are many AI algorithms that are open-source and free. It’s the cost of the data required.”
Pitfalls to avoid
Besides the already-mentioned technical risks linked with data availability and quality, some of the biggest pitfalls to avoid are forgetting the human factor, setting expectations too high and biased data. AI can automate many repetitive tasks and deliver significant return on investment, but it cannot replace people entirely. This means supply chain organizations still require a combination of automation and human interaction, which introduces an increased likelihood of human error.
Likewise, ML and optimization models are written by people, who might introduce biases and errors. And even when that is not the case, unpredictable and unprecedented situations occur. Maltaverne also flags the dangers of biased data: “Relying on bad data is dangerous, as AI and ML will act on it and make things worse.”
Supply chain challenges that AI and ML can help solve
Business cases for investment in supply chain AI and ML should be framed around targeted solutions to your organization’s pain points. Rigonat lists several challenges that AI and ML can advance, including production planning, demand forecasting, inventory management, routing, dynamic pricing, finding the best suppliers, maintenance planning, facility location (depots and warehouses), fraud detection and quality control in the production line.
“Just as business rules engines have yielded great results in fraud detection and back-office tasks automation in the finance sector, they could do the same for supply chain,” she says. “AI and ML can also be successfully used for programming customer service chatbots to improve and automate time-consuming tasks that are still handled manually.”
Moreover, better buying — in particular, ESG-related decision-making — is currently very difficult due to a lack of data. “Every decision is a complex one because it involves finding the right balance and juggling trade-offs with limited visibility,” Maltaverne says. “AI can plug into data lakes across silos and provide the recommendations needed for better buying, especially when it is combined with the potential of real-time internet-of-things data and the trust generated by blockchain.”
For Maltaverne, the key challenge that AI and ML can help with is VUCA (volatility, uncertainty, complexity and ambiguity). He notes that most organizations do not have a proper risk-management framework in place, which becomes starkly obvious when an incident occurs. “Take the Ever Given in the Suez Canal: How long did it take for organizations to understand the blockage’s impacts and put actions in place to adapt to it? Having an always-on, AI control tower connecting an organization to partner ecosystems and the outside world makes an enormous difference.”
