Title 03_P05 final-18-27.pdf ✅

20 HỘI THẢO KHOA HỌC QUỐC GIA VỀ LOGISTICS VÀ QUẢN LÝ CHUỖI CUNG ỨNG VIỆT NAM LẦN THỨ 4 (CLSCM-2024) indicated that some AI procurement software enables purchase orders to be automatically reviewed and approved [9]. This application allows firms to be more proactive in their inventory management systems and mitigate the risk of stockout. d) Real-time traceability First, the real-time tracking capabilities of AI can help firms identify items that are moving slowly, alert them to potential stockouts, and recommend necessary adjustments to their inventory levels [10]. In addition, businesses can increase transparency, improve recall management, and develop consumer trust by tracking and documenting the movement of goods [11]. 3.2. Impacts of AI-based inventory management on business AI-based inventory management is attributed to remarkable revenue growth. Take the example of Whole Foods. Amazon acquired Whole Foods and then opened its own Amazon Go employee-free automated retail store chain in 2017. Since then, they have used AI to understand customer behavior in both in-store and e-commerce channels, optimize inventory strategies, and improve the consumer shopping experience. As a result, the Company can enhance its revenue remarkably. A study shows that the application of AI will likely help companies generate about $1,000 billion in new profits by 2030, equivalent to about 10% of the total profits they expect to achieve at that time [12]. AI-based inventory management can significantly drive down supply chain costs. An AI risk-based system using manufacturing and inventory data to simulate, predict, evaluate, and optimize logistical and supply bottlenecks can save 15% on operations and maintenance costs [13]. An actual case study of a retail chain in China, Auchan Retail, shows that AI- driven inventory optimization algorithms have resulted in a 7% reduction in inventory [14]. 3.3. Conditions for adopting AI in inventory management a) Business needs and goals Only firms having complex inventory management systems with enormous stock keep units (SKU) should consider AI. Moreover, businesses should have a clear goal when trying to adopt AI in their operations. b) Data quality and integration Ensuring the quality and integration of data is a major condition. For AI systems to produce exact results, complete and accurate data is required. The precision of demand forecasting is a critical component of AI-enhanced JIT inventory systems [11]. c) Integration with existing systems The firm should ensure that the AI system integrates seamlessly with the existing inventory management systems for a smooth workflow [15]. The companies should establish an environment where tech and the business team can collaborate to create and implement AI solutions. It is necessary to involve business teams to solicit feedback and test prototypes of AI models [13]. d) Cost and resource requirement The investment in AI technology, including hardware, software, and skilled personnel, can be substantial and costly. The application of AI can be divided into three phases and each phase is implied with associated cost. To implement AI, some assessments and competence tests must be completed during the pre-adoption phase. In the adoption phase, employees must possess the requisite knowledge and skills. An investment in human capital is needed for effective AI applications. After that, maintaining and updating AI technology was a crucial due to the continuous development of the company [16]. 3.4. Criteria for a successful application of AI in inventory management According to new research, successful AI programs require AI alignment. In this framework, a successful AI system needs to meet three criteria. a) Scientific consistency First, scientific consistency aims to minimize the gap between reality and the AI model. To reach scientific consistency, developers compare the results of AI programs with empirical evidence. If discrepancies, firms should upgrade the model by adjusting data, features, algorithms, or domain knowledge. b) Application consistency To reach application consistency, an AI model also needs to achieve targets and prevent unforeseen effects. Before deploying AI, it's important to define its scope. This involves considering potential impacts and unintended consequences. Based on this, the firm can identify the necessary adjustments. c) Stakeholder consistency The ideal AI solution aligns with the needs of all
21 HỘI THẢO KHOA HỌC QUỐC GIA VỀ LOGISTICS VÀ QUẢN LÝ CHUỖI CUNG ỨNG VIỆT NAM LẦN THỨ 4 (CLSCM-2024) stakeholders, from company leaders to everyday users. This means the program delivers clear value that everyone involved understands and ultimately profits from [17]. 4. Analysis of Walmart’s AI-powered inventory management 4.1. AI practices in Walmart’s inventory management a) Demand forecasting and inventory planning Regarding the process of building an AI/ML- powered inventory management system, Walmart starts with a foundation of data and business restrictions to generate a universe of potential ML models. The Firm uses web searches, page views, and historical data, such as previous sales, to fine-tune ML models. Additionally, to predict demand and bottlenecks in delivery, the system also considers "future data" like weather patterns, macroeconomic trends, and local demography. Moreover, Walmart’s AI/ML engines' can "forget" the abnormalities, preventing the influence of one-time deviations into subsequent inventory management. The establishment of AI model provides result based on its algorithms. Walmar also have their employees make final decision thanks to their experiences throughout their long career path. Moreover, during usage, AI models are continuosly trained to improve results accuracy [18]. b) Inventory control Walmart uses integrated information systems with the support of RFID (Radio Frequency Identification) tags for inventory management. These tags help to enable real-time tracking of individual items. The data would be fed into AI systems to provide real-time inventory levels across stores and distribution centers. This allows for proactive management and identification of stockouts or surpluses. In 2019, Walmart tested a system called Intelligent Retail Lab (IRL). Thanks to this system, Walmart can enable staff to know about the exact situation of each shelf and determine when to replenish products [19]. c) Automated reordering AI algorithms analyze this data stream in real time to identify replenishment needs. Based on the demand forecast and current inventory levels, the system determines when and how much needs to be restocked. When necessary, the system automatically generates purchase orders and sends them to suppliers. The current Walmart’s real-time replenishment system can handle billions of communications from 100 million SKUs in less than three hours [20]. 4.2. Lessons from Walmart’s AI application in inventory management experiences a) Firmly identify business needs and goals Other firms should evaluate their existing inventory management system to answer the question if they need AI or not. Walmart is a retail giant with more than 10,000 stores and 400 million SKUs [21]. A powerful inventory management system is necessary for Walmart to optimize stock levels, ensure product availability, and enhance customer experience. b) Ensure data quality and integration The company should provide a massive amount of high-quality data for the AI work. In 2020, the strategy to merge grocery pickup into its primary shopping app helped the firm’s database more comprehensive to carry out in-depth analysis [22]. Additionally, the input quality has an enormous influence on the reliability of the model’s results. However, Walmart is pursuing real-time supply chain visibility. Therefore, it can track inventory throughout its supply chain [19]. c) Ensure integration with existing systems It will be meaningless if the AI application cannot align with the current system. AI is the new recent technology that Walmart is using in its inventory management system but information technology (IT) in its operations has been applied for a long time. Walmart’s inventory management is designed to always be ready for technology. Therefore, when applying such new technology as AI, it is not difficult to align between the tools and the existing system. d) Prepare cost and resource requirement To successfully leverage AI in inventory management, the company should prepare resources, in which financial and human are the most important ones. Walmart proactively invested in AI and technology development over the past few years [23]. Human resources are also an essential factor for a successful application of AI. Since 2015, Walmart's employment for AI-related positions has doubled [24]. Additionally, the Company insists that Walmart is a “people-led and tech-powered” business because its associates possess wonderful intuition to make the right decision [20]. 5. Suggestions for AI application in inventory management across Vietnamese retail enterprises 5.1. Overview of the retail industry in Vietnam