Expert Column Framework and Use Cases for the Adoption of Generative AI (GenAI) in Logistics

Definition of GenAI and Increased Use of GenAI by Companies Generative AI (or Generative AI system) is also called ‘Generative AI or GenAI’ in English, and the word ‘Generative’ means ‘capable of producing or originating.’ There is no strict definition of GenAI, but it can be said to be ‘AIs that can generate a variety of content’ or ‘AIs that have the learning ability to generate a variety of content.’ These AIs can create various types of content such as text, images, music, video, and program code, and new content is transformed into creativity based on the existing data.

In the first half of 2023, when ChatGPT had over 100 million users, the term ‘Foundation Model’ began to become a buzzword among corporate executives. The foundation model (meaning basic model when translated into Korean) has various definitions, but the theoretical rationale is that it can demonstrate tremendous productivity if used within the scope of the company or government. This term was created and popularized by the Stanford Institute for Human-Centered Artificial Intelligence’s (HAI) Center for Research on Foundation Models (CRFM). CRFM is much cheaper when it comes to adjusting existing base models or directly using generative AI such as ChatGPT in specific use cases. Therefore, if you use it without developing it yourself, you can easily and quickly create a foundation for utilization. As the theoretical basis for this was established, it provoked tremendous interest from executives of global companies.

As it was theoretically proven that ChatGPTs can be a groundbreaking tool for improving customized productivity and creativity with low cost and less effort for corporate use, a craze for full-fledged adoption began in March 2023.

McKinsey’s August 2023 ‘Global Survey on the current state of Generative AI’ report showed that interest in GenAIs among global companies is explosively increasing. One- third of respondents said their organizations were regularly using GenAIs in at least one business function. Recent developments have seen AI rapidly emerge as a focus for CEOs and CIOs, with 60% of respondents saying that advances in AI will cause their organizations to increase their investments in AI overall. They responded that ChatGPT would increase the company's productivity by more than 20% and that most of the organization's members would need to be retrained. Strategy Framework for Application and Adoption of GenAI in Logistics The author would like to present the ‘2x2.4 Framework’ as an important tool to systematically overcome various challenges that may arise when introducing GenAI in the logistics industry. This framework is designed to help companies simply classify complex tasks by application area, select generative AI-applicable tasks, and determine the scope of the project. This framework was published in my book, 'Making ChatGPT A Secret Weapon of Innovation (2024, Knowledge Nomad).’

This 2x2.4 framework is divided into two parts. These are ‘2x2 matrix of application areas’ and ‘4 types of materialization’ for unit project development. The ‘2x2 matrix of application areas’ consist of four quadrants based on two main axes. The horizontal axis is divided into 'internal operation task' and 'customer service', and the vertical axis is divided into 'improvement of corporate work' and 'creation of new tasks'. This helps clarify the organization's various strategic priorities. ‘Four types of materialization’ are the criteria for judging the extent to which it will be linked and applied with GenAI after the target task is selected. These four types are the standard created by analyzing application cases of generative AI by global companies throughout a year in 2023. These two were combined and named ‘2x2.4 Framework’.

2x2 Matrix and Cases in Logistics Quadrant 1 : External Customer Service – Improving Existing Tasks
The first quadrant, [External customer service - Improvement of existing tasks], includes the following services: a delivery tracking and notification service that notifies customers of delivery status in real time and provides an ETA(Estimated Time of Arrival), an automated customer inquiry response service that provides automatic responses to delivery-related inquiries using an AI ChatBot, and a customized delivery option suggestion service that recommends the optimal delivery option by analyzing customer's purchase history and delivery data.

Logistics Optimization at ‘Sydney Duty Free’ in Australia
Following the COVID-19 pandemic, Sydney duty free shops have been facing new challenges with rapidly increasing number of tourists. Manpower shortages and rising labor costs in Australia have placed a significant burden on duty-free shop operations, which has seriously affected customer service and operational efficiency. Amid this situation, Sydney Duty Free decided to innovate its business operations by introducing AI technology, especially OpenAI's ChatGPT.

Optimize Operation and Logistics - ChatGPT played an important role in inventory management and logistics processes. Traditional inventory management and logistics were primarily manual, which often resulted in inventory shortages or overstock problems. This was a major cause of higher operating costs and lower customer satisfaction. ChatGPT's predictive analysis feature estimated demand based on sales data and automated inventory management. Through this, it effectively solved problems of inventory overstock and shortage, and greatly improved operational efficiency.

Quadrant 2: External Customer Service – New Task Creation
[External Customer Service – New Task Creation] in the Quadrant 2 includes: Customer Support Agent, which predicts and responds preemptively to customers’ demand with AI; AI-based Product Recommend System which recommends new products by analyzing purchase patterns; and Personalized Marketing Service, which offers customized promotion and discount information.

Stich Fix: Product Recommendation
The clothes of Stitch Fix, an online personal styling service company, are constantly changing and increasing. To create new services, it introduced ChatGPT and applied it to various fields. As Stich Fix wanted to show its customers what was timely and on-trend, it had to create 13 million new outfit combinations every day. For this purpose, it created an OCM (Outfit Creation Model) based on ChatGPT. ChatGPT trained millions of outfits created by stylists so that the OCM can collect outfit suggestions considering real-time inventories, customers’ preferences and purchase history. Through this process, recommendations for each customer is gaining positive attention.

Quadrant 3: Internal Operation – Improvement of Existing Tasks
The Quadrant 3 [Internal Operation – Improvement of Existing Tasks] includes: Inventory Management Optimization, optimizing inventories using AI and preventing inventory shortage or overstock; Logistic Route Optimization, finding the best way of delivery based on AI and reducing costs; and Advanced Data Analysis Service, analyzing logistics data, providing insights to improve efficiency and outcomes.

Samsung SDS: Route Optimization and Data Analysis using Cello Square
It analyzes complex routes in the logistics network and recommends optimal transportation routes, cutting costs and enhancing efficiency. In addition, it can be utilized as a tool that not only provides insights to evaluate and improve performance by analyzing logistics operation data (delivery time, cost, error rate, etc.) but automatically creates and manages logistics-related documents such as invoices and waybills.

In May 2024, Samsung SDS announced that it is working to advance logistics services by implementing hyper-automation by combining ChatGPT technology with Cello Square, an existing logistics service platform. Cello Square is a digital logistics platform where customers can use all logistics services, from quotation to reservation, transportation, tracking, and settlement. Samsung SDS launched the “Cello Square Logistics Service,” an interactive logistics service, on the GPT Store.

Cello Square is decreasing supply chain risks based on data collection. We collect 60,000 pieces of news related to global supply chain issues on a daily basis. Afterward, it figures out how much those risks are related to the actual logistics using machine learning. GenAI, which has learned about 20,000 past supply chain risk issues, selects an average of 70 high-risk news articles per day and delivers them to managers. The time it takes to establish a plan was also shortened from one day to two hours, accelerating the speed of response.

Quadrant 4: Internal Operation Job – New Task Creation
The Quadrant 4 [Internal Operation – New Task Creation] is comprised of Robot-based Logistics Automation to classify and move products in warehouses; Demand Prediction Analysis which expects future demand and plans accordingly; and Intelligent Supply Chain Management Service for Establishing AI System, which monitors and manage in real-time.

Maersk: Prediction Analysis
Maersk, the world's largest container shipping company, is attempting to use GenAI to perform predictive maintenance on its ships. Analyzing a ship's sensor data to forecast maintenance needs reduces downtime and improves ship reliability. GenAI can dynamically adjust routes and prices based on real-time traffic data, fuel costs and delivery schedules, ensuring on-time delivery and reducing fuel consumption How Method To Connect With GenAI and Practices How have companies used GenAI such as ChatGPT in connection with individual tasks? If we classify the cases observed so far, they can be roughly divided into four patterns.

1. Role Defining Type
This is a method of enhancing the organization's work efficiency by independently utilizing features provided by GenAI, rather than connecting it with the company's existing system or data-generating AI. It demonstrates the fastest outcomes by using it independently for daily tasks such as writing emails, summarizing meeting minutes, translating, and correcting spelling.

2. Function-Linkage Utiliziation Type
This is a type that promotes customer experience by linking GenAI conversation feature with existing systems. A representative example is identifying consumer needs through conversational bots and then recommending optimal products and services.

3. Fine-tuning Integration Type
This refers to the use of GenAI as a customized intelligent bot by training companies’ professional data. ChatGPT learns financial market analysis reports, product manuals of manufacturers, medial thesis of hospitals, etc., coming up with replies at the expert level.

Finance Area: Morgan Stanley
In September 2023, Morgan Stanley announced that its GenAI assistant was “fully operational” for financial advisors and supporting staff. Morgan Stanley made a strategic decision to adopt GenAI technologies not only to utilize the latest technologies but also to offer better services to customers and maximize the efficiency of internal operations.

Morgan Stanley adopted a ChatGPT-based chatbot to promptly respond to customer’s inquiries. This chatbot can answer inquiries regarding account information, investment portfolios, market trends and complex financial terms with simple explanations. This improves customer satisfaction as they are able to get the information they need whenever they require it.

GenAI was also used to derive important insights while analyzing a massive amount of data and to write research reports based on this. Morgan Stanley's analysts are now able to make more accurate and faster market predictions based on data provided by GenAI. Currently, Morgan Stanley is using ChatGPT to develop an additional tool called Debrief, which automatically summarizes customer meetings and generates follow-up emails.

4. Multimodal Convergence Type
It is a linking method that combines two or more GenAIs with different characteristics to innovate the process of target tasks or create a new service. It combines AIs which create different content, like combining voices and images or texts and videos Things to Consider for Successful Adoption of GenAI Many companies are not clear on where to start or what areas to focus on when adopting GenAI. Last but not least, companies must take the following five things into account to successfully introduce GenAI.

1) Estimate effects of utilization and set goals
The success of using GenAI depends on applying GenAI first to tasks that will see the biggest effects. Rather than proceeding to use it without a policy or strategy, it is important to figure out how it can be matched with tasks and set goals on how it can contribute to efficiency and productivity.

2) Select highly effective areas and design processes for usage
GenAI is very useful in creating content based on vast amounts of data, but it cannot guarantee it gives right answers to complex questions, and AI hallucinations surely exist. Therefore, a step of reviewing and choosing which task/purpose to apply and which approach/scope/tool to use is crucial as this step could determine the outcome of the GenAI.

3) Development and adoption of agile approaches
GenAI should be conducted with an agile approach (create, twist, and evolve) that involves data preparation, fine-tuning of models, and monitoring and improvement, not by adopting and using an information system. Specifically, it is recommended to take an agile development approach, a repeating two-week cycle of quickly building a prototype, actually using it, reviewing it and revising it.

4) Manage risks from a system and a rule
The biggest reason companies hesitate to utilize GenAI is due to a concern about risks such as leakage of confidential information or copyright infringement. It is possible to manage and minimize risks by establishing a system that does not learn input data and making operation rules, such as scope of use and handling of confidential data.

5) Improve employees’ AI utilization literacy
A characteristic of GenAI is that it provides an output through a dialogue with users, so it is considerably influenced by the users' literacy. Thus, training programs and hands-on training are required to create an environment where employees understand the basic knowledge of GenAI, appropriate usage and accompanying risks, and utilize it in a responsible manner.

▶ This content is protected by the Copyright Act, and the copyright is held by the contributor.
▶ This content is prohibited from secondary processing and commercial use without prior consent.