Generative AI for supply chain management

Insights

• The Infosys Generative AI Radar, North America, reported user experience and personalization as the most positive use cases of generative AI in supply chain management.
• This emphasizes generative AI’s near-term value from external-facing processes like customer service, in contrast to internal ones that focus on operations and efficiency.
• Recent disruptions caused business leaders to prioritize resilience, sustainability, and localization for supply chain management.
• Generative AI offers myriad applications to improve supply chain operations.
• Organizations must explore potential generative AI use cases that enhance efficiency, increase customer satisfaction, and drive non-linear growth.

The recent Infosys Generative AI Radar study for North America revealed several insights for supply chain management. SCM is a board room topic, alongside chief information security officers and cyber security experts as senior decision makers. However, the champion for implementing generative AI in enterprise supply chains remains unclear. Supply chain ranks lowest in establishing business value through use cases.

However, user experience and personalization emerge as the most positive use cases of generative AI among supply chain and logistics companies (Figure 1). This emphasizes generative AI’s near-term potential value from external-facing processes like customer service, in contrast to internal ones that focus on operations and efficiency.

Figure 1. Generative AI creates the greatest impact on user experience for North American supply chains

Source: Infosys Knowledge Institute

The shift in priorities due to post-pandemic disruptions and events such as the Ukraine war significantly hindered generative AI use cases in supply chain management. Traditionally, supply chains focused on speed, cost, and quality metrics. These disruptions caused business leaders to prioritize resilience to recover from unforeseen shocks, sustainability for least carbon emission, localization to mitigate risks and diversify supplier base, and agility for timely and rational decision making (see Figure 2).

Figure 2. Traditional vs. contemporary digital supply chains

Source: ASCM’s SCOR model, Infosys Knowledge Institute

We see the emergence of a digital supply chain, with information as the driver for decision making. AI and digital twins enable stress testing to identify weak areas which require investment and management attention. Sourcing critical parts such as semiconductor chips remains a weak link, as an example. A spike during the pandemic in high-end chips for consumer goods affected other industries that used low margin chips. Timely information flow among stakeholders preempts disruptions, enabling AI to convert data into actionable insights that drive decision making.

Generative AI in supply chain management holds immense potential for resiliency and efficiency. Organizations must explore potential use cases that enhance efficiency, increase customer satisfaction, and drive non-linear growth. Projections suggest global generative AI supply chain investment to reach $13 billion (about $40 per person in the US) by 2032 at a CAGR of 46% from $301 million in 2022. For North America, the recent Infosys generative AI survey that involved over 1,000 respondents estimated the investment to double in the next 12 months to around $70 million.

Opportunities and challenges

Generative AI offers myriad applications to improve supply chain operations. Readily deployable use cases include automating customer services, synthesizing and automating document creation (contracts, logistics and custom papers), demand forecasting, and predictive maintenance. Beyond new content creation, generative AI understands patterns from existing content and competitive insights, according to a recent article in Harvard Business Review.

Despite the enthusiasm around generative AI’s potential advantages, supply chains present multiple challenges, including data security, privacy, and access to publicly available AI tools within corporate IT landscapes. These concerns prompt firms to develop generative AI solutions in-house or with vendors, potentially missing proven off-the-shelf toolkits offered by open models.

For skill development, organizations tend to partner with vendors to tap into their talent pool; for internal talent development, they upskill existing employees familiar with industry rules and regulations. So far, external talent recruitment has been a lower priority. The industry is tentatively ready to adapt and implement generative AI but may gather momentum with sufficient leadership direction and support.

Gartner recently stated that 80% of the global supply chain is not incorporated into current digital decision models. The digital-to-reality gap will continue to hamper supply chain performance objectives until technology investments are complemented with decision support, says Gartner.

The Association for Supply Chain Management (ASCM) identifies AI and ML (Machine Learning) among the top 10 trends in supply chain management. However, it attributes the delay in generative AI adoption to the intricacies of supply chains and the necessity for company-specific model training.

Simple, adoptable use cases for productivity include document drafting (emails, manuals, contracts), predictive analysis for risk management and mitigation, and automated customized content generation for marketing campaigns. Also, conversational AI is well-placed to manage customer service and query resolutions.

Large language models for conversational AI

Conversational AI is an evolved form of chatbots, combining natural language processing (NLP), large language models (LLMs), and machine learning to enable dynamic conversations. As LLMs advance, conversational AI becomes highly sought after across sectors for its capacity to provide real-time interactive use cases that enhance user experience.

Supply chain management is a regulated industry with country-specific legal frameworks. It can benefit with LLMs built using existing knowledge and kept up to date. These LLMs answer queries, classify new documents, and create new documents such as contracts using prompt engineering. Siemens implemented an NLP tool in partnership with Infosys to classify tax documents, keeping Siemens tax consultants updated on taxation changes, classifying data, and summarizing tax-related conversations. This LLM (Large Language Model) model, built using GPT-3 and Microsoft Azure OpenAI Service, facilitates tax-related knowledge classification and automated newsletter generation.

While supply chains address issues like data security and the complexities of integrating digital technologies, it presents extraordinary opportunity for significant growth and improved efficiency. “Introducing AI into supply chains is not just an improvement of current processes, it represents a complete overhaul of how decisions are made, and operations are conducted.” says Dr Mukesh Kumar, associate professor, Institute for Manufacturing, Department of Engineering, University of Cambridge. The challenge is to move through this rapidly evolving landscape by strategically developing skills, adopting modern technologies, and striking a balance between custom-built solutions and open AI tools.

Generative AI provides insights by analyzing trends and predicting disruptions, optimizes logistics and inventory through predictive analytics, and automates operational tasks to improve efficiency. While generative AI may enhance agility and competitiveness, its use in supply chain strategic decision making remains unexplored. However, efforts are underway to advance AI utilization in this arena.

Advanced use cases across the value chain

The following use cases represent generative AI applications in supply chain management, subject to data to build models that accurately represent current ways of working.

• Production planning and control. This fundamental topic drives the rest of the supply network. Convenience store giant 7-Eleven aims to deploy generative AI for product planning. 7-Eleven plans to generate texts and images through AI for new products based on analysis of store sales data and consumer feedback through social media. The new system is expected to accelerate cycle time by 10% and improve product quality.
• Sourcing and procurement. Generative AI transforms the supply chain sourcing and procurement process. It handles purchase orders, negotiates deals, performs supplier selections, and assists in contract preparations. For example, Walmart deployed generative AI to automate supplier negotiations. Walmart shares its budget and requirements with the software, which then interacts with human sellers to finalize each deal.
• Manufacturing. Generative AI gains momentum on the factory floor, with automated code generation for programmable machines and robots. Among its many manufacturing applications, Siemens plans to generate programmable logic controller (PLC) code through natural language inputs to reduce time and errors.
• Delivery and reverse logistics. Generative AI offers powerful sustainability and cost-saving benefits in supply chain processes, particularly in delivery logistics. Its applications include trade optimization, trade network redesign, and last mile delivery enhancement through logistics and route optimization. Infosys and Google Cloud generative AI recently helped a consumer goods company successfully launch an AI twin to conduct real-time planning of marketing spend, promotion, and product supply across markets. DHL, a premier logistics and supply chain company, pioneers generative AI adoption in its delivery process and logistics through route optimization and supply chain monitoring.
• Returns. Generative AI aids supply chain companies in tracking and accounting carbon emissions, specifically Scope 3 emissions and circularity metrics which have been difficult through traditional methods.

Supply chains are the delivery network of today’s global economy. They move physical goods and data simultaneously, worldwide. Unfortunately, this also elevates cybersecurity and data privacy as top concerns for generative AI adoption. Governance mechanisms and security systems design are persistent priorities to adopt generative AI for supply chain management.

“Generative AI has significant potential for value in the decision-focused areas of supply chain,” says Robert Winans, vice president, head of supply chain, Panasonic Energy North America. It’s inevitable that generative AI will greatly enhance areas like raw materials planning, sourcing, procurement, and logistics.

Recommendations

We recommend four practices for business leaders to accelerate their generative AI journey, in stages of increasing complexity:

• Standardize and simplify before digitizing. This is crucial for supply chain management, which cuts across organizations, geographies, languages, business units, functions, and even industries. Simple business processes create low-risk success while controlling costs and timelines, prioritizing off-the-shelf solutions over customization.
• Start with simple, high demand use cases to understand generative AI implementation and achieve early successes. Customer complaint handling and related support functions are attractive initial candidates.
• Collaboration is required for all supply chain initiatives, and generative AI is no different. Collaboration should occur across units and beyond organizational boundaries, as upstream and downstream partners act at their respective nodes and possess valuable information that could enhance models.
• After straightforward initial use cases, companies must make bold decisions and rapidly adopt generative AI due to its rapid growth and evolution. The gap between early adopters and those planning adoption will widen quickly.

To stay competitive, firms should investigate advanced implementations at each supply chain stage. Supply chain management faces significant volatility, uncertainty, complexity, and ambiguity. Timely data flow and AI developing valuable insights offers firms the tools needed to address these challenges.