Streamlining Distribution Routes Using the Language Model of Artificial Intelligence

Abstract

This article addresses the use of artificial intelligence for the needs of effective, sustainable development in logistics and its components. The subject of this article is to highlight the possibility of processing optimization methods using an artificial intelligence module. The goal is to determine whether the AI module can replicate the same, or at least have a similar result, as the traditional optimization methods used in practice. The challenge involves constantly identifying reserves in already highly sophisticated micro-logistics systems using modern commercial means of artificial intelligence. Applying artificial intelligence to elements of a company’s micro-logistics model is a new approach. This article aims to determine whether artificial intelligence can reduce costs through calculations in a specific area defined for it. By optimizing distribution routes using ChatGPT-3.5, we significantly reduced the total distance traveled, leading to substantial savings in transportation costs. This optimization led to a significant improvement in the efficiency of logistic processes and considerable cost savings. This result demonstrates that artificial intelligence can be an effective tool for solving complex logistic tasks. The possibilities of effectively sustainable logistics development with the help of artificial intelligence lie not only in the quality of the achieved outputs but also in the speed of the calculations and the procedures for solving defined project tasks. It follows from this definition that artificial intelligence will continue to play an essential role in the defined field of logistics in the future.

1. Introduction

Approximately 67% of the world’s population, which equates to 5.4 billion people, are now online [1]. Globally, up to 4.88 billion people use a mobile phone [2]. According to the latest data, the number of people worldwide who are not connected to the internet decreased to an estimated 5.5 billion in 2024 [3].

In this context, the internet represents inherent access to information, job opportunities, and education. People who have meaningful access may be included. This becomes increasingly important as technologies like artificial intelligence (AI) become more prevalent in our daily lives [4]. Demis Hassabis, co-founder and CEO of Google DeepMind, believes that AI has the potential to become one of the most important and beneficial technologies ever invented [5].

Sustainable logistics is closely intertwined with the concept of artificial intelligence, because AI technologies enhance the efficiency and effectiveness of logistics operations, enabling better resource management, reducing waste, and optimizing processes to minimize environmental impact. Sustainable logistics is an approach to managing supply chains that minimize environmental and social impacts while maintaining economic efficiency. It involves reducing emissions, efficiently using energy and materials, and minimizing waste. The economic aspect includes optimizing processes to lower costs and increase productivity. Social sustainability focuses on fair working conditions and adherence to ethical standards. Innovations and modern technologies, such as intelligent transportation systems and automation, play a key role in improving efficiency. This approach integrates environmental, economic, and social factors into a cohesive system [6].

The societal impact of AI is further explored by Yuzhou QIAN and other authors in their publication “Societal Impacts of Artificial Intelligence: Ethical, Legal, and Governance Issues” [7]. Generative language models of artificial intelligence, such as ChatGPT, have existed for some time. By early October 2023, approximately 18% of the Slovak population had tried AI modules (such as ChatGPT) at least once. This amounts to about 780,000 people from the Slovak population [8]. With the growing influence of AI and its models, it is essential to monitor and analyze its development to harness its advantages while minimizing its risks effectively.

ChatGPT is a robust NLP model capable of understanding and generating natural language for various applications, including text creation, language comprehension, and interactive programs. The future of ChatGPT and the consideration of its various possibilities for study and development, such as energy efficiency, cybersecurity, expanding its usability to other technologies (robotics and computer vision), enhancing communication between humans and AI, and bridging the technological gap, are discussed by authors Sukhpal Singh Gill and Rupinder Kaur in their publication titled “ChatGPT: Vision and Challenges” [9]. One of the goals of ChatGPT is to transform how people interact with computers and automate tasks by leveraging the potential of generative AI. The ability of ChatGPT to converse like a real human being is one of its most impressive features [10].

Due to its promising applications, the ChatGPT language model is becoming well known for everyday use. Suppose we want to connect the mentioned AI model with the industry. In that case, it is undoubtedly important to view it through the lens of Industry 4.0, which promises to revolutionize processes that help advance productivity and business efficiency. At the same time, within the industry, its integration can reduce costs and improve decision-making processes across the entire supply chain management system. ChatGPT can automatically process and analyze large amounts of data from various sources, aiding in the planning and optimization of the supply chain. With its ability to generate solutions and responses to questions, ChatGPT can quickly react to changes and risks in the supply chain. Its capacity to provide personalized recommendations and support enhances the efficiency and productivity of processes in supply chain management. These points are discussed in more detail by the authors in the publication “Smart Supply Chain Management in Industry 4.0: The Review, Research Agenda and Strategies in North America” [10].

2. Review of the Subject Literature

2.1. The Growing Influence of Artificial Intelligence (AI) in Logistics

AI tools and digital platforms, or at least tools and platforms claiming “intelligent status”, have become indispensable to business organizations and society over the past decade. This stems from the ability of AI algorithms to automate business processes, extract knowledge from big data, provide predictions and recommendations, and offer superior analytical and computational capabilities compared to human beings [11]. Various forms of AI, such as robotic process automation (e.g., cobots in warehouses), computer vision techniques, speech recognition, machine and deep learning algorithms, and natural language processing, have created numerous opportunities and unique possibilities for organizations to reengineer business processes and functions, focusing on innovating business models and offerings in the consumer space (such as agile and data-driven decision-making, project management, and strategic planning for critical productivity indicators) [12,13].

The COVID-19 pandemic has further highlighted our reliance on technology availability and the importance of Europe not depending on systems and solutions from other regions. The Digital Europe program provides strategic funding to address these challenges, focusing on projects in five key capacity areas: supercomputing, artificial intelligence, cybersecurity, advanced digital skills, and ensuring the widespread use of digital technologies across the economy and society, including through digital technologies [14].

Most sectors, including logistics, have been transformed by developing new technologies like the AI above. Autonomous vehicles, warehouse automation, predictive analytics, and intelligent roads are examples of technologies that are becoming the new norm today. “Digitalization is crucial for logistics companies. The days when warehouses operated in a ‘paper and pencil’ style are long gone. Through digitalization, companies can significantly streamline processes, reduce costs, and manage overall logistics operations”, says Patrik Jokl, founder of IPRICE RECARE [15].

AI is utilized in numerous situations, such as listening to music via a voice assistant, using a mobile camera in portrait mode when the correct parameters are set, listening to music selected based on algorithms that identify consumer trends, using a route planner that chooses the best path to the destination, or in connection with an email spam filter that automatically performs this classification [16].

By collecting and analyzing data such as traffic conditions and transport capacities, AI offers a unique opportunity to plan routes more efficiently for future trips, as the data can be intelligently combined and optimal connections planned, which, among other things, reduces empty runs and increases the efficiency of available resources.

AI optimizes much more today. For example, shortly before the arrival of a truck, employees can prepare the appropriate loading dock, speed up the loading and unloading process, and promptly dispatch the truck. Thus, AI is already helping to streamline the organization of supply chains. Automation of these work processes is one of the main tasks in logistics. By using intelligent applications, processes can be increasingly automated. This saves time and costs, and manual work is minimized [17].

There are already numerous examples of the integration of automation and AI. Shippers can use logistics systems to find a suitable transport partner and integrate digital functions into the supply chain. Real-time tracking enables quick responses to issues in the supply chain, whether they involve traffic jams, accidents, or insufficient supply. In the “last mile” area, comprehensive tests are already being conducted today, such as with transport drones, autonomous vehicles, or 3D printing products for individual or urgent orders. These innovations, especially for end consumers, could become faster and less dependent on traditional transportation methods [17].

Google DeepMind uses AI to predict the structure of over 2 million new materials. In an article published in Nature, the AI company states that nearly 400,000 of its hypothetical material designs will soon be possible to produce in a lab. This research’s potential applications include producing more efficient batteries, solar panels, and computer chips. The discovery and synthesis of new materials can be costly and time-consuming. An example is lithium–ion batteries, currently used to power everything from phones and laptops to electric vehicles. It took approximately two decades of research before they became commercially available [18].

DeepMind’s AI was trained using a database of existing research on approximately 50,000 known materials from the Materials Project, an international research group established at Lawrence Berkeley National Laboratory in 2011. The company states that it will now share its data with the research community to accelerate further groundbreaking discoveries in materials science. “The industry typically avoids risk when it comes to increasing costs, and it usually takes a long time for new materials to become cost-effective”, says Christine Pershon, the project’s director. “If we can shorten this period even a little, it will be a real breakthrough” [18].

In the context of using artificial intelligence for logistical purposes, the available literature is still relatively limited. Existing studies often focused on theoretical models and simulations, while practical applications in the real world are less documented. This creates an opportunity for further research and experimentation, which could yield new insights and contribute to the broader application of artificial intelligence in logistics processes.

The objectives of the research were focused on the utilization of artificial intelligence to achieve efficient and sustainable development in the field of logistics. One of the main goals was to explore whether an AI module, specifically ChatGPT, could replace traditional optimization methods used in logistical processes. For the article, it was important to determine whether ChatGPT could effectively identify reserves in already highly sophisticated micro-logistics systems through modern commercial AI tools.

Another goal was to verify if the application of artificial intelligence could lead to cost reductions in a specific area of logistics, namely through the optimization of distribution routes. By optimizing these routes, the aim was to see if improvements in the efficiency of logistics processes and cost savings could be achieved. The research also aimed to demonstrate that ChatGPT could be an effective tool for solving complex logistical tasks.

In this way, the possibilities of AI for quick and precise calculations necessary for solving defined project tasks were tested. It was interesting to observe how ChatGPT could improve the overall efficiency and bring significant cost savings to logistics. The successful optimization of distribution routes using ChatGPT confirmed that artificial intelligence could be an asset in logistical processes. Such an approach enabled more efficient resource utilization and cost reduction, which are crucial for sustainable development in this field.

The research also had a secondary objective to pave the way for further applications of artificial intelligence in logistics, showing that these technologies will play an important role in the future. Overall, it is essential to demonstrate that ChatGPT and similar AI tools can bring innovative solutions and improve the current logistics systems.

2.2. Revolutionary Innovations in Supply Chain Management: Artificial Intelligence Leading the Transformation

The dawn of generative AI has the potential to transform logistics and supply chain management radically. The boom in e-commerce, ongoing labor shortages, emerging geopolitical events, and the constant threat of disruptions place new demands on global supply chains. Organizations are investing in technologies, automation, and robotics to enhance efficiency, reduce costs, increase visibility, and foster more agile and responsive supply chains [19].

The emergence of AI has immense potential to revolutionize the way new products are created within supply chains, enabling companies to improve market research, accelerate ideas and design, streamline collaboration, optimize demand forecasting, and enhance quality control. These improvements lead to greater efficiency, shorter time to market, and increased customer satisfaction. However, it is important to note that collaborative AI is not meant to replace human creativity and expertise. Integrating AI into product creation processes within supply chains represents a significant opportunity to drive innovation, maintain competitiveness, and deliver top quality products in a dynamic market environment [20].

Today’s supply chains are more complex due to material shortages, climate disasters, and geopolitical tensions, making the logistics process even more challenging [21].

Supply chain disruptions are seen as a severe risk by 39.1% of logistics professionals, though cybercrime and a shortage of skilled workers are considered even more dangerous [22]. A survey by ZipDo revealed that 37% of supply chain leaders are already using or planning to use AI within 24 months. Accenture is expanding its partnership with SAP SE to create a nerve center using AI to enhance supply chain resilience and sustainability, featuring the Cosmo Tech Supply Chain Vulnerability Scan to simulate potential outcomes [21].

AI simulation is a new generative AI that can automatically run thousands of simulations to help supply chain managers identify vulnerabilities that will have a significant impact on their business and pinpoint weak spots that could cause future problems. It assesses the impact of supply chain disruptions and the cascading effect on every supplier and route while running scenarios based on changes made in these areas. Unlike other AI models based on historical data, AI simulation uses synthetic data to create thousands of potential future scenarios [21]. Generative AI accelerates discovery, making discovery a competitive advantage [23].

The evolving landscape of supply chain management also highlights the transformational potential of integrating intelligent software. The pandemic brought challenges, disrupted production, led to shutdowns, and caused delays and increased costs in global transportation. Irregular changes in consumer behavior led to demand fluctuations, affecting different industries in various ways. The crisis emphasized the need for digital transformation, remote work, and adapting technologies in supply chain management. Financial pressures underscored the importance of robust risk management, leading to a renewed focus on building resilient and flexible supply chains. Companies are now investing in technology, diversifying suppliers, and reevaluating inventory strategies. Intelligent software enhances decision-making and risk management and facilitates collaboration across the supply chain. For example, during sudden demand shifts due to lockdowns, the software quickly analyzed data, allowing for immediate adjustments in inventory, production, and distribution [24].

The emergence of ChatGPT has heightened interest in the AI chatbot market. Microsoft’s USD 10 billion investment in OpenAI, along with new services from Google, Meta, Baidu, and others, has intensified this trend [25,26]. For example, Baidu introduced “Wenxin Yiyan” or “Ernie Bot”, and Naver launched its AI tool HyperCLOVAX. Microsoft also introduced Dynamics 365 Copilot for efficient supply chain management, which automatically generates reports and alerts suppliers to potential risks [27].

On 1 December 2022, OpenAI launched a conversational chatbot named ChatGPT, which could produce moderately complex text on a wide range of topics. ChatGPT was based on the GPT architecture, a relatively new form of AI that predicted outputs based on user inputs. This approach, generative AI, allows anyone to open ChatGPT and chat with the AI on relatively complex topics. Moreover, ChatGPT can generate output for any user regardless of their training or background. This is the first reason business approaches to AI changed rapidly: ChatGPT demonstrated that AI systems could be accessible and usable by anyone, not just a select few AI engineers [28].

In March 2023, OpenAI unveiled its new GPT architecture iteration, GPT-4. This model exhibits reasoning capabilities that surpass the average human performance across almost all evaluation metrics and exceed the top 5% of human performance in most assessment parameters [29,30]. When GPT-4 was endowed with the ability to write, interpret, and execute code, it could conduct quite complex data analysis and visualization [29]. When the standard model is used simply as a chatbot, it allows any person to produce relatively sophisticated material, with some even suggesting that GPT-4 could replace human research [31]. GPT-4 is, at the time of writing, the most potent generative AI model, but other models are also being developed concurrently by companies like Google. This competitive race to produce ever more powerful AI systems, which are widely available, ensures that increasingly capable AI systems will be in the hands of professionals sooner rather than later [28].

According to ChatGPT’s CEO Sam Altman, the potential impact of these new AI variants could represent a breakthrough in the field of technology, with significant implications for employment, stakeholder relationships, business models, and academic research. The introduction of more advanced and powerful generative AI tools into the AI market after the launch of ChatGPT has heightened the “AI arms race”, creating ongoing uncertainty for workers, expanding opportunities for businesses, and simultaneously increasing the risks associated with their use [32].

2.3. Direct Applications of AI in Supply Chain Management (SCM)

• Enhanced market research and consumer insights

AI is transforming market research by providing valuable data and analyses for product development. AI algorithms analyze consumer data, such as social media trends, reviews, and purchasing behavior, allowing companies to better understand preferences and trends. This enables firms to develop products better tailored to market needs, increasing the likelihood of success without wasting time and resources on trial and error [20]. Generative AI can also automate customer interactions, such as order updates and handling complaints, which improves customer satisfaction and service quality [25].

• Streamlining supply chain collaboration

AI can enhance collaboration in the supply chain by enabling efficient sharing of information and documents in real time. Generative AI can support sustainability by providing a platform for collaboration between designers, engineers, and suppliers, reducing delays and misunderstandings. This approach speeds up decision-making and improves information flow throughout the product development process [20,33].

• Predictive analytics for demand forecasting

AI can revolutionize demand forecasting by analyzing historical data and market trends. Accurate demand forecasts enable companies to optimize production, inventory management, and supply chain operations, helping to minimize stock and enhance customer satisfaction. This way, AI provides valuable insights for better decision-making regarding production and distribution [20].

• Coordinating distribution, transportation, and logistics

Generative AI can design effective distribution and logistics strategies, considering factors such as cost, service levels, weather conditions, and environmental concerns. AI can also create contingency plans for disruptions like traffic jams. By analyzing real-time data, AI can suggest the most efficient transportation routes [25].

• Quality control and product testing

AI improves quality control and product testing by analyzing data from various sources, including customer feedback and sensor data. Automating inspection processes reduces errors and increases efficiency. AI can also simulate product testing scenarios, helping to predict and optimize the product performance and lifespan [20].

• Supplier evaluation and sourcing

Generative AI offers a revolutionary approach to supplier evaluation by quickly analyzing large volumes of data from potential suppliers and considering various parameters such as cost, quality, and sustainability. This approach allows for the optimization of the supplier portfolio [25].

AI significantly enhances various aspects of the supply chain, from market research and demand forecasting to quality control and supplier evaluation, ultimately driving greater efficiency, accuracy, and customer satisfaction [20,25,33].

3. Materials and Methods

This article addresses the issue of continuously improving efficiency in logistics through the application of artificial intelligence (AI). AI brings revolutionary capabilities for optimizing logistics processes, ranging from demand forecasting to delivery route optimization. The study focuses on leveraging these technologies to tackle specific challenges within the logistics sector.

In recent years, large language models (LLMs) have revolutionized natural language processing (NLP) with their exceptional abilities to understand and generate text similar to human language. These models, which employ deep learning techniques and are trained on vast amounts of data, have led to significant advancements in various NLP applications, such as machine translation, question answering, text summarization, and conversational systems [5,9] ChatGPT, developed by OpenAI, has become a prominent example of these models, gaining widespread attention from both the public and researchers due to its ability to generate coherent and contextually appropriate responses. These models have become an effective solution to the limitations of traditional NLP techniques, using deep learning and extensive text data to learn linguistic patterns [27,32].

Large language models’ ability to generate contextually relevant text is a key factor in their success. These models learn from the provided context in input prompts and generate responses that consider previous information, leading to more accurate and appropriate answers. The availability of models like ChatGPT has opened new possibilities for NLP applications, including integration into chatbots, virtual assistants, and customer service systems, enabling more human-like interactions. They enhance tasks such as language translation, summarization, and information retrieval with faster and more precise results [34].

Since the introduction of ChatGPT, AI chatbots have become popular due to their ability to handle a wide array of tasks that can be beneficial both personally and professionally. The global recognition of ChatGPT has driven competitors to develop diverse versions, leading to a wide array of market options with different strengths, use cases, and difficulty levels [34].

It is important to understand how ChatGPT works for the study. The ChatGPT algorithm is based on the transformer architecture, specifically the GPT (Generative Pre-trained Transformer) model. This architecture utilizes a series of transformer layers designed to process data sequences, such as text [34].

Brief Description of the ChatGPT Algorithm:

• Tokenization of Input
  The input text is split into tokens and then encoded into a vector of numerical values. This tokenization allows the model to represent text in a form that it can quickly process.
• Positional Embedding
  Information about the position of each token in the sequence is added to the tokens. This enables the model to recognize the text’s patterns and relationships between tokens.
• Transformer Layers
  The text is processed through a series of transformer layers, allowing the model to understand the context and relationships between tokens in the input. These layers perform operations like multi-head self-attention and feedforward neural network layers.
• Output Generation
  The model generates output text based on the processed input. This can be an answer to a question, a continuation of a conversation, or any other textual output.
• Output Decoding
  The numerical values generated by the model are decoded back into text form, which is then presented to the user as the final response.
• Feedforward Network
  At the end of the model, a feedforward network layer processes the output from the transformer layers and performs the final transformations needed to generate the output text [34,35].

This process is repeated for each input sequence provided to the model, allowing ChatGPT to generate contextually meaningful responses to various questions and prompts.

Understanding the model’s internal processes and decision-making mechanisms also helps recognize it as a fundamentally important tool. The flowchart can be viewed as a tool that visualizes the processes and decision-making procedures the model uses to generate responses and interact with users. It offers valuable insight into how information is processed and how the model produces contextually relevant answers (Figure 1).

• Data Acquisition and Processing

GPT-3 is trained on vast textual data from the internet and other sources. These data are processed and used to train the model.

• Model Training

The data are utilized to train a neural network with many parameters. The training process involves techniques like deep learning and reinforcement learning.

• Model Creation

After successful training, the GPT-3 model is created to generate text based on the input provided.

• Input Data

GPT-3 receives input data in the form of text sequences, which the model processes.

• Text Generation

Based on the input data, output text is generated. GPT-3 attempts to predict the most likely continuation of the text based on what has been seen so far.

• Text Output

The generated text is presented to the user or another system that utilizes it.

• Feedback and Further Training

In some cases, feedback from users or other methods can be used to improve the model’s performance. This feedback can be used to update the model and for further training.

• Iteration:

Training and using the model can be repeated to achieve better quality and accuracy in the generated text [34,35,36].

ChatGPT can be a critical interface tool for addressing logistical costs and enhancing institutional efficiency. It can assist with numerous tasks, especially those that involve high repetition and redundancy. However, it is important to remember that ChatGPT is not a substitute for human intelligence. Human intelligence is rooted in biological processes and includes understanding contexts, interpreting meanings, and creating connections.

On the other hand, ChatGPT has excelled in generating human-like text and found applications across various industries. However, it has its limitations: despite them, it has a clear future and impact in the field of NLP, where it continues to evolve and likely surpass its current constraints. There is a growing trend in using language models in areas such as customer service, content creation, and translation. Its scalability is another advantage, as it is capable of processing large volumes of data without delay, making it ideal for companies with high demand flows. ChatGPT and large language models are increasingly utilized in diverse sectors such as finance, medicine, education, and marketing, contributing to the future of these fields [37].

Simultaneously, it has proven a valuable tool for efficient, low-cost content creation across various industries. It is a fast tool for generating materials and revolutionizing content production. For instance, it generates investment and analytical reports in the financial industry. In the media industry, it enables the automatic creation of news and articles, thereby shortening the creative cycle. ChatGPT has broad application possibilities and can be a vital tool in domains requiring natural language processing and understanding. With continuous technological development, its impact and utilization are expected to expand further [38].

Overall, ChatGPT represents an essential step in the development of artificial intelligence and holds significant potential across various fields, including logistics. Its ability to communicate and provide information can benefit people, yet the ethical and technical aspects of its usage must be carefully considered. Over time, and with further development, these benefits could deepen while minimizing drawbacks, potentially opening new possibilities for its application [36].

ChatGPT utilized the Traveling Salesman Problem (TSP) method and clustering to optimize the distribution routes. It used tools like Google Maps, ViaMichelin, and RouteXL to obtain precise distances and travel times between cities. These tools allowed it to verify and optimize routes, focusing on minimizing distances and travel times. It input the data into RouteXL to find the most efficient route for each circuit. Excel was used to record and compare the original and optimized data. It implemented the TSP algorithm to find the shortest route and clustering to group locations. Subsequently, it compared the optimized routes with the original ones. It found that the optimized routes resulted in savings in distances and costs. Using these methods achieved improved distribution efficiency.

4. Results

One of the goals of this article was to verify whether ChatGPT can tackle specific logistics tasks—specifically, optimizing transportation routes. Using advanced algorithms and data analysis, ChatGPT contributes to increasing efficiency and reducing costs in logistic processes. Initially, a logistics task was formulated to determine if ChatGPT could process it. The assigned task was the optimization of distribution routes within specific areas.

During the optimization of distribution routes, the following procedure was used with ChatGPT: First, ChatGPT was asked “Can you optimize the routes in the table based on the data I sent you to reduce the overall distances in the distribution routes?”

ChatGPT responded “Yes, I can help you optimize the routes based on the data you send me. I need the following information from you:

• A list of addresses or points to visit: Ideally in the order they are currently visited.
• The distances or time between these points: This could be in the form of a distance matrix (for example, a table where the rows and columns are addressed and the cells contain the distances or time).
• The starting and ending point: If the route needs to start and end at a specific location.
• Additional constraints or preferences: Any specific requirements such as maximum travel time, the maximum number of points on one route, time windows for visiting each point, etc.

Please send me this data, and I will commence the analysis and route optimization”.

Based on the provided input data, ChatGPT analyzes the problem and selects the optimal method for its solution. The input data provided to ChatGPT for solving the assigned task are in

Table 1. Input data—distribution routes.

Distribution Area	Start	Direction	End	Total km	Transportation Cost
1		Rožňava–Dobšiná–Jelšava–Plešivec		315.4	485.72
2		Poprad–Svit		179.5	276.43
3		Kežmarok–Spišská Belá–Stará Ľubovňa		180.4	277.82
4		Lemešany–Sabinov–Lipany		94.4	145.38
5		Hanušovce nad Topľou–Holčíkovce–Vranov nad Topľou		130.5	200.97
6		Humenné–Snina–Medzilaborce		238.3	366.98
7	Prešov	Michalovce–Sobrance	Prešov	221	340.34
8		Trhovište–Palín		205.4	316.32
9		Sečovce–Trebišov		159	244.86
10		Borša–Kráľovský Chlmec–Veľké Kapušany		267.6	412.10
11		Spišská Nová Ves–Levoča–Spišské Vlachy		149.6	230.38
12		Gelnica–Krompachy		97.6	150.30
13		Košice–Bidovce–Čaňa–Moldava nad Bodvou		171.1	263.49
14		Raslavice–Bardejov–Svidník–Stropkov–Giraltovce		142.3	219.14
			Total (KM) (EUR Diesel—1.54 EUR 5/24)	2552.1	3930.23

.

Based on the input data from Table 1, ChatGPT was tasked with optimizing the distribution routes using an appropriate method. ChatGPT’s response was as follows: “Based on the provided data, we can optimize the routes to reduce the total number of kilometers driven. We can achieve optimization by reordering visited locations and possibly splitting the routes. We will follow these steps:

• Identify the points in each route and their distances.
• Optimize the order of locations within individual routes (for example, using the Traveling Salesman Problem–TSP algorithm).
• Consider possible regrouping of points between routes (if time and capacity constraints allow)”.

ChatGPT chose the TSP method to complete the task. The author directed ChatGPT to combine the TSP method with the clustering method for better results. ChatGPT evaluated this instruction as optimal. The TSP (Traveling Salesperson Problem) algorithm is handy for planning distribution routes, as it minimizes the total distance traveled by vehicles, thereby reducing fuel costs, travel time, and overall logistic efficiency. The algorithm employs various methods, including heuristic and metaheuristic approaches, enabling it to find optimal or near-optimal solutions even with many locations, as was the case with the task we assigned. In this instance, ChatGPT used the TSP algorithm to analyze and optimize the distribution routes, leading to more efficient resource utilization and cost savings in transportation.

Based on the data provided in Table 1, ChatGPT has proposed optimizing the routes to reduce the overall distance traveled. The optimization was proposed by clustering distribution routes based on geographical proximity from the 14 circuits provided as input data. ChatGPT proposed computing optimized routes for the following

Table 2. Optimized distribution routes using ChatGPT.

Cluster	Start	Direction	End	Total km	Transportation Cost
1		Kežmarok, Spišská Belá, Stará Ľubovňa, Sabinov, Lipany, Lemešany		223.9	344.81
2		Rožňava, Dobšiná, Jelšava, Plešivec, Poprad, Svit		328	505.12
3		Michalovce, Sobrance, Trhovište, Palín, Sečovce, Trebišov		262	403.48
4	Prešov	Hanušovce nad Topľou, Holčíkovce, Vranov nad Topľou, Humenné, Snina, Medzilaborce	Prešov	249	383.46
5		Spišská Nová Ves, Levoča, Spišské Vlachy, Gelnica, Krompachy, Košice, Čaňa, Moldava nad Bodvou		292	449.68
6		Raslavice, Bardejov, Svidník, Stropkov, Giraltovce		139	214.06
7		Borša, Kráľovský Chlmec, Veľké Kapušany		232	357.28
		The individual routes present deviations in kilometers.	Total (KM) (EUR Diesel—1.54 EUR 5/24)	1725.9	2657.87

:

After entering the input data, ChatGPT optimized the distribution routes by creating seven clusters. This clustering of distribution routes enabled more efficient travel and reduced the total distance. Combining clustering with the TSP algorithm ensured overall improvement in the distribution logistics, reducing the distribution route from the previous 2552.1 km to 1725.9 km. Through this optimization, ChatGPT helped reduce transportation costs from 3930.23 EUR to 2657.89 EUR.

Following the optimization that reduced the distribution routes and transportation costs, the length of each distribution route was recalculated. Upon verification, it was found that the distances of the distribution routes showed deviations compared to the distances available on the internet, such as through Google Maps. Therefore, additional calculations were necessary to determine the specific deviations caused by ChatGPT’s calculations.

In

Table 3. Optimized distribution routes using ChatGPT and our own calculations using Google Maps.

Cluster	Start	Direction	End	Total km	Transportation Cost
1		Kežmarok, Spišská Belá, Stará Ľubovňa, Sabinov, Lipany, Lemešany		238.9	367.44
2		Rožňava, Dobšiná, Jelšava, Plešivec, Poprad, Svit		374.7	577.04
3		Michalovce, Sobrance, Trhovište, Palín, Sečovce, Trebišov		279.6	430.58
4	Prešov	Hanušovce nad Topľou, Holčíkovce, Vranov nad Topľou, Humenné, Snina, Medzilaborce	Prešov	250.2	385.31
5		Spišská Nová Ves, Levoča, Spišské Vlachy, Gelnica, Krompachy, Košice, Čaňa, Moldava nad Bodvou		334.9	515.75
6		Raslavice, Bardejov, Svidník, Stropkov, Giraltovce		138.3	212.98
7		Borša, Kráľovský Chlmec, Veľké Kapušany		246.5	379.61
		Recalculated using Google Maps	Total (KM) (EUR Diesel—1.54 EUR 5/24)	1862.8	2868.71

, it shows the final optimization of all the routes. All distances of individual clusters were verified and optimized using Google Maps for accuracy. Several discrepancies were identified through Google Maps, amounting to a total difference of 136.9 km compared to the previous calculations. These discrepancies translated to 210.82 EUR in total transportation costs. With the assistance of ChatGPT using clustering and the TSP algorithm, the total distance of the distribution routes was reduced from 2552.1 km to 1862.8 km. The optimization of all the routes resulted in total transportation cost savings from 3930.23 EUR to 2868.71 EUR.

5. Discussion

In his study on optimizing distribution, author Martin Straka [39] used the ratio index method to concentrate on reducing or optimizing distribution routes and the related distribution costs. The ratio index method (RIM) is used to optimize distribution routes, assessing and enhancing the efficiency of transporting goods or services. This method involves calculating ratio indices for different routes based on criteria such as costs, time, distance, or fuel consumption. These indices can identify the most efficient routes and optimize logistical operations.

Before applying the ratio index method, the distribution routes covered a total distance of 2552.1 km. After implementing the ratio index method, the distribution routes were reduced to 1391.5 km, which represents a reduction of 45.50%. The AI module optimized the distribution routes using the same input data used with the ratio index method.

ChatGPT has significant potential as an auxiliary tool within logistics. For this article, it was verified whether ChatGPT can handle specific logistics tasks, including optimizing distribution routes and reducing costs. Based on the provided data, ChatGPT optimized distribution routes by clustering them according to geographic proximity, reducing the total kilometers traveled.

Also, ChatGPT uses the TSP method. The Traveling Salesman Problem (TSP) is a well-known algorithmic problem in the fields of computer science and operations research, focused on finding the shortest possible route that visits each point in a set of cities exactly once. It is highly relevant to logistics, where it helps optimize routes and reduce transportation costs. The Traveling Salesman Problem (TSP) often requires adjustments to accommodate additional constraints, such as time windows or vehicle capacity. Technological advances and the development of heuristic and metaheuristic algorithms have made finding practical solutions more feasible. This problem is combinatorial and NP-hard, meaning the number of possible solutions grows exponentially with the number of cities, which is why approximation algorithms are often used to find effective solutions [40,41,42].

ChatGPT proposed route optimization based on these methods to reduce the distance traveled. By clustering distribution routes from 14 circuits, ChatGPT created seven clusters, enabling more efficient travel and reducing the total distances. Combining clustering and the TSP algorithm improved logistics distribution, reducing the distribution route from 2552.1 km to 1725.9 km and cutting transportation costs from 3930.23 EUR to 2657.89 EUR.

This optimization resulted in significant savings and enhanced efficiency in logistic processes. Despite its advantages, ChatGPT has limitations, such as inaccuracies in specific calculations or mathematical tasks, as highlighted during the computations. After the optimization, the length of each distribution route was recalculated, revealing discrepancies compared to Google Maps data. These discrepancies amounted to a difference of 136.9 km in total distance and 210.82 EUR in costs. These drawbacks and potential errors suggest that users should be aware of ChatGPT’s limitations and supplement its responses with additional sources and fact-checking.

ChatGPT demonstrated that the total distance of the distribution routes could be reduced to 1862.8 km and transportation costs to 2868.71 EUR. It has been proven that ChatGPT can process the logistics tasks assigned to it. This optimization resulted in significant savings and improved efficiency in logistics processes (

Table 4. Comparison of the results.

Type of Route	Total km	Transportation Cost
Original	2552.1	3930.23
Optimized	1862.8	2868.7
Difference	689.3	1061.52

).

This optimization resulted in significant savings and enhanced efficiency in logistics processes. Despite its advantages, ChatGPT has limitations, such as inaccuracies in specific calculations or mathematical tasks, as highlighted during the computations. After the optimization, the length of each distribution route was recalculated, revealing discrepancies compared to Google Maps data. These discrepancies amounted to a difference of 136.9 km in the total distance and 210.82 EUR in costs. These drawbacks and potential errors suggest that users should be aware of ChatGPT’s limitations and supplement its responses with additional sources and fact-checking.

Although ChatGPT can handle basic calculations and logical reasoning, it may struggle with complex mathematical computations or detailed tasks, leading to inaccuracies. Many authors in their articles emphasized the need to verify generated responses, as ChatGPT is limited in its ability to provide answers for which it was not trained. The authors also pointed out that ChatGPT lacks personal experience and context-specific understanding, which further limits its ability to provide relevant answers. It is important to keep in mind that ChatGPT is still in the developmental phase and continues to improve and evolve. These drawbacks and potential errors suggest that users should be aware of ChatGPT’s limitations and supplement its responses with additional sources and fact-checking [26,43].

The provided calculation demonstrates that ChatGPT could be utilized even more broadly and effectively. It could be applied in real time to monitor and adjust distribution routes based on current traffic conditions, weather, or unexpected events. Additionally, it could be integrated into inventory management systems to forecast demand and optimize supply based on historical data and market trends. ChatGPT could also assist in warehouse operations planning, supply chain management, and improving communication among various stakeholders in logistics processes [44].

Research shows that deploying AI chatbots enhances supply chain visibility and supports innovative capabilities. Small- and medium-sized manufacturing enterprises can gain greater agility and responsiveness in their operations thanks to these chatbots, which help track inventory, plan deliveries, and analyze demand. The speed of operations can also be demonstrated by the calculations conducted as part of the distribution route optimization. AI chatbots are transforming logistics and supply chain management by providing integrated solutions for optimizing operations and improving the prediction of consumer preferences [45].

Other studies have revealed that integrating digital technologies into industrial processes, such as data analysis through AI, is most influential in improving sustainability and economic performance [46]. Therefore, digitizing logistics, aided by ChatGPT, is becoming an urgent requirement for companies [47].

With the continuous development and enhancement of AI technologies, ChatGPT’s potential in logistics will continue to expand, bringing new opportunities and improved solutions to this critical area.

6. Conclusions

Language models like ChatGPT, harnessing artificial intelligence and machine learning, can address simple logistic tasks such as optimizing distribution routes. Therefore, artificial intelligence is becoming increasingly important in the field of sustainable logistics efficiency. These tools enhance the accuracy and speed of solving logistics challenges, improve efficiency, and reduce costs. Specifically, the use of AI in micro-logistics systems can result in significant savings in transportation costs and enhance the overall efficiency of logistical operations.

This article focused on whether ChatGPT can handle basic logistics tasks and help reduce costs. Through its proposed calculation, ChatGPT aimed to optimize distribution routes to minimize the overall travel distances and transportation costs. ChatGPT reduced transportation costs and the total distance traveled by 27.01%. ChatGPT, when applied to the optimization of distribution routes, did not yield better results than those obtained using the ratio index method. While processing the distribution optimization, it could not account for all the requirements and factors influencing it. However, the optimization achieved by ChatGPT can still be considered successful, given that it reduced the distribution routes and transportation costs.

Employing advanced technologies to optimize logistics processes provides companies with a competitive edge and offers opportunities for long-term sustainability in a rapidly evolving digital environment. Companies must rethink their business models, processes, and methods to ensure continuous and reliable supply chain operations. In this context, digitization and innovative technologies have become essential tools that enable the necessary and expected changes in this field.

On the other hand, it is important not to mention only the advantages of ChatGPT. The use of ChatGPT in supply chain management can present several challenges and limitations, as highlighted in this article. The quality of the input data is crucial; incomplete or inaccurate data can lead to incorrect results. The model may have limited expertise in specific industries, which can result in less accurate or ineffective decisions in certain specialized areas of the supply chain. Data security and confidentiality are also key concerns when using AI. Security when using ChatGPT and AI is a critical aspect that deserves special attention. It is important to ensure that personal and sensitive data processed by AI systems are protected from unauthorized access or misuse. Implementing strict data protection measures, such as encryption and stringent access controls, can help minimize risks. Another important aspect is ensuring transparency in how data are processed and what algorithms are used for decision-making. Users should be informed about how the AI system operates and what data are used to train the models. Transparency can enhance trust and enable better understanding and control of the system. Dependence on AI may reduce human oversight of processes. Additionally, the use of AI raises ethical dilemmas, such as its impact on jobs and responsibility for AI decisions, and it is necessary to consider the transparency and fairness of algorithms.

In the context of logistics optimization, artificial intelligence can have a significant impact on society and practical applications in the field. Artificial intelligence can significantly enhance the efficiency of logistics processes by reducing transportation costs, optimizing distribution routes, and increasing overall productivity. These improvements can lead to substantial economic savings for businesses, enabling them to lower their costs and improve customer service.

From a societal perspective, the widespread use of artificial intelligence in logistics may affect the job market. Some routine and repetitive tasks could be automated, leading to the replacement of specific jobs. However, new job opportunities may arise in areas such as AI technology development and maintenance, as well as analytical and strategic roles necessary to operate these systems effectively.

Comparing this to research in other fields, such as the automotive industry, where autonomous vehicles and advanced navigation systems are already creating fundamental changes, artificial intelligence in logistics could produce similar transformative effects. Research in the automotive sector shows that automation can reduce transportation costs and increase road safety, outcomes that could similarly be achieved in logistics through AI. For example, in his study, Voss addressed the problem of stochastic vehicle routing with uncertainty in the number of available vehicles. Within this context, the generative artificial intelligence system ChatGPT was examined. It investigated whether ChatGPT could provide meaningful answers and effectively address theoretical problems, emphasizing the need for careful use and consideration of ethical issues. The research revealed that, despite current limitations, ChatGPT, when integrated with other systems, can significantly enhance its practical utility in areas such as logistics, indicating the need for further development and applications [48].

Economic systems may also be impacted, as the streamlining of logistics processes could change how products and services are distributed, potentially leading to lower prices and improved product availability for end consumers. Ultimately, the development and implementation of artificial intelligence in logistics bring practical benefits in terms of savings and efficiency and create new challenges and opportunities that require adaptation and innovation across various areas of society and the economy.

As part of the continuation of this research, there are plans to delve deeper into artificial intelligence modules and their potential in logistics. The goal is to expand the understanding of artificial intelligence’s practical applications and develop new methods for solving complex logistics tasks.