Breaking Boundaries: Exploring Generative AI’s Impact on Supply Chains

Breaking Boundaries: Exploring Generative AI’s Impact on Supply Chains

Supply chains encompass many interconnected activities, from procurement, production, and inventory management, to logistics and distribution. These activities involve numerous stakeholders, such as suppliers, manufacturers, distributors, and retailers. Efficiently managing these processes is vital for companies to meet customer demands, minimize costs, and maintain a competitive edge. With its ability to derive insights from vast amounts of data and derive insights, generative AI has emerged as a valuable tool to optimize supply chain operations.

‘Traditional’ AI vs Generative AI

Before delving into generative AI, it is necessary to understand where it fits concerning AI generally. Traditional AI operates with relatively simple models (say, thousands of parameters) on relatively small datasets (say, thousands of images) to produce relatively simple outputs (image shows a fault or not). Generative AI is all about scale. Generative AI can use models with billions or even trillions of parameters trained on massive datasets (think the size of the internet), enabling it to produce increasingly complex outputs such as speech, text, and videos. The AI utilizes unsupervised and semi-supervised machine learning to understand statistical patterns and structures from enormous datasets. It can then generate new content using prompts that resemble the original training data.

The seemingly miraculous capabilities of generative AI systems like ChatGPT (large language models, or LLMs, generating novel text and code), Dall-E (generating novel images), and Alphafold2 (protein structure prediction) come from exploiting this scale. AI-based tools already help automate routine tasks and assist decision-making processes across various sectors. AI models have grown tenfold, representing a step-change in AI capabilities, and creating new use cases across the supply chain.

Generative AI and LLMs have potential benefits but face challenges such as data quality across large data sets and ethical implications with bias, fairness, and transparency. Despite these challenges, organizations are already using generative AI to manage unstructured datasets.

The Role of AI in Supply Chains

Before delving further into generative AI, let’s consider the broader implications of AI use in supply chains. Through machine learning and data analytics, companies can gain valuable insights from extensive datasets, supporting decision-making. By leveraging the power of AI, organizations can optimize efficiency, cost-effectiveness, and overall customer satisfaction.

Supply chain data can be challenging to manage due to its unstructured or semi-structured nature and the need to integrate new data sources. However, generative AI can extract insights from documents like waybills, quality reports, and customer feedback and feed it into AI models to optimize various aspects of the supply chain based on historical supply chain data. Furthermore, generative AI can generate text, images, and code, potentially speeding up administrative tasks.

In IDC’s recent Future Enterprise Resiliency & Spending Survey (Wave 2, IDC, April 2023), an average of 40% of supply chain organizations already invest in generative AI, with knowledge management as the top focus for generative AI applications.

Figure 1: What’s your organization’s current approach to Generative AI, and which use cases have the most promise for your organization?

Source: IDC, Future Enterprise Resiliency & Spending Survey – Wave 2, April 2023

In a supply chain context, knowledge management applications refer to the use of generative AI to capture, store, organize, and leverage large datasets and insights generated throughout the supply chain processes. Manufacturers, retailers, and logistics organizations can leverage generative AI-powered knowledge management systems to optimize operations and make informed decisions. Table 1 details some examples of areas of application of generative AI for different supply chain participants.

Table 1: Uses of Generative AI

Source: IDC, 2023

To use generative AI effectively, organizations should identify where it can add value, evaluate their infrastructure for compatibility, and prepare their workforce. To ensure successful integration, they should create a plan with timelines, resources, and evaluation metrics (refer Figure 2).

Figure 2: A technology adoption roadmap for transportation and logistics companies considering generative AI.

Source: IDC, 2023

Challenges and Limitations

To fully utilize the potential of generative AI, we must acknowledge and address the many obstacles that hinder its widespread implementation. These challenges require a strategic and systematic approach to overcome them. Only by doing so can we unlock the full benefits of this innovative technology.

  • Data Quality and Quantity: To create effective generative AI, high-quality training data is necessary. However, in some fields, obtaining labelled or annotated data can be difficult. It’s important to note that the AI model’s understanding of the world is based on the training data it receives. Therefore, biased or unfair outcomes may result from biased training data, requiring careful handling and pre-processing. Additionally, since few companies have access to large datasets, partnering with other companies to share quality data may be necessary to maximize the potential of generative AI.
  • Interpretability and Explainability: Generative AI models, such as deep neural networks, are complex and black-box in nature, making it difficult to interpret the decision-making process, which can hinder trust and adoption. In high-stakes domains where understanding the reasoning behind AI-generated outputs is crucial, this is particularly important.
  • Ethical and Legal Considerations: AI-generated content raises ethical concerns around privacy, bias, and accountability. Organizations must ensure ethical use, avoid biases, and comply with data protection laws.
  • Generalization and Robustness: Generative AI models struggle with new or challenging situations, resulting in unpredictable results and unreliability. Ensuring the robustness and reliability of generative AI models across diverse contexts remains a challenge.
  • Computational Resources and Efficiency: Training and deploying generative AI models can be computationally intensive and require significant resources, such as high-performance, large-scale computing infrastructure, limiting accessibility and scalability for organizations with limited resources and impacting sustainability. Many firms will have to use pre-trained models and tune them to local more specific datasets and tasks.
  • Fine-grained Control and Creativity: Balancing the fine-grained control and the desired level of creativity in generative AI models can be challenging. While fine-grained control ensures adherence to specific requirements, it may restrict the generation of novel and creative outputs.
  • Legal and Intellectual Property Issues: AI-generated content can raise issues surrounding ownership and intellectual property rights, especially if the AI model has been trained on copyrighted or proprietary data.
  • Human-AI Collaboration and Trust: Integrating generative AI into existing workflows and ensuring effective collaboration between humans and AI systems can be challenging. Building trust among stakeholders, clear communication, and understanding of AI capabilities and limitations is crucial for successful adoption. Job loss concerns can also hinder progress.

The use of generative AI in the logistics and supply chain industry has the potential to serve as a gateway technology, catalyzing the broader adoption of AI across the entire supply chain ecosystem. Generative AI can help industries solve challenges and improve operational efficiency. By using historical data and external factors, accurate models can optimize supply chain operations, demonstrating the potential of generative AI and encouraging its adoption in other areas such as supply chain visibility, process automation, risk management, and route optimization. As more organizations witness its positive impact, they become more open to exploring other AI applications, leading to wider integration in their supply chain.

About the Author

Stephanie Krishnan
Associate Vice President, IDC Asia/Pacific

Stephanie Krishnan is an Associate VP responsible for producing, developing, and growing the IDC Manufacturing, Retail, and Energy Insights programs in Asia/Pacific. Within Manufacturing Insights, Stephanie conducts supply chain, sustainability, and Industry 4.0 research that supports clients in areas such as global sourcing (profitable proximity and sustainable outcomes), risk management and resilience, transportation, logistics, warehousing, digital transformation, and automation.