The digital economy has risen to become the engine of the global economy, with AI technology being its key driving force. China's digital economy has grown rapidly, with technological innovation continuously advancing AI. The deep integration of the digital economy and AI provides rich data resources for AI technology, promoting its continuous upgrading. The introduction of large AI models creates more opportunities for the global digital economy and drives industries towards high-quality development. At the same time, AI technology has also triggered profound changes in the field of enterprise services, improving efficiency, reducing costs, providing personalized services, and creating new business value and economic benefits for enterprises. In the future, AI will penetrate more widely into various fields of enterprise services, including marketing, operations, customer service, and digital employees, achieving comprehensive intelligent improvement.
On September 27, Li Qing, Director of Frost & Sullivan Greater China, officially released the '2023 China AI Technology Transformation Enterprise Service White Paper' at the 2023 Second New Investment Expo and the 17th Frost & Sullivan Global Growth, Innovation and Leadership Summit - Digital Economy and Industrial Integration Sub-forum. The paper analyzes the application and transformation of AI technology in marketing, operations, and customer service sectors from the perspective of digital and intelligent transformation.
Li Qing, Director of Frost & Sullivan Greater China
This white paper focuses on the research on transformative applications brought about by AI-related technologies in the broad field of enterprise services. By surveying the AI technology transformation enterprise service industry chain, core technologies, and application scenarios in China from 2022 to 2023, it explores how AI technology in China has led to development and application upgrades and extensions in enterprise services. It also analyzes industry scale, implementation applications, and future trends, providing an understanding and interpretation of AI-transformed enterprise service scenario applications.
Industrial digitization is the engine of digital economic growth. Through deep integration with innovative digital technologies, its scale and depth have rapidly increased, making it a core role in the digital economy.
"
Industrial digitization continues to dominate the digital economy, closely integrating with constantly innovative digital technologies. Its scale and depth are rapidly increasing, becoming an engine for digital economic growth. On the stage of the digital economy, industrial digitization firmly occupies a dominant position, intricately intertwined with the innovation of digital technologies. With the rapid progress of digital technologies, the internet, big data, artificial intelligence, etc., are being closely integrated with the real economy, making industrial digitization even more crucial in driving digital economic growth.
In 2022, the scale of China's digital industrialization reached 8.8 trillion yuan, accounting for 18.3% of the digital economy. This figure not only witnessed a rapid quantitative growth but also highlighted that industrial digitization is moving towards a path of quality improvement. At the same time, in 2022, the scale of industrial digitization climbed to 39.3 trillion yuan, accounting for as high as 81.7% of the digital economy. It is expected that by 2025, the overall scale of the digital economy will reach 55.7 trillion yuan. These data not only highlight the vigorous progress of industrial digitization in its in-depth development but also emphasize its core position in the engine of the digital economy. It not only maintains a leading role in scale expansion but also plays an increasingly important role in quality upgrading and deep innovation, thus demonstrating an increasingly significant impact on overall economic growth.
AI empowers enterprise service architecture, which is divided into four sections: the foundational layer, model layer, AIGC tool layer, and enterprise service layer. AI large models provide a wide range of products and services for different industries in the field of enterprise services.
"
The AI-powered enterprise service architecture includes the foundational layer, model layer, capability layer, and enterprise service layer. The core technology layer integrates AI technology and large models to provide products and services for various industries. Currently, the commercialization of AI-driven enterprise services is in its initial stages. Large technology companies are committed to building multimodal large models to achieve Model as a Service (MaaS), while vertical industries focus on fine-tuning and training specific domains, providing customized products and services, and promoting the deep integration and innovation of AI technology and large models.
AI infrastructure is supported by high-quality networks and centered around data resources, algorithm frameworks, and computing power. It leverages open platforms as its main driving force, providing the underlying architecture for intelligent services to the public in the long term.
"
AI computing power includes AI chips, intelligent computing centers, AI cloud centers, etc., providing strong computing support for the development of artificial intelligence technology and industries. Currently, artificial intelligence technologies represented by deep learning require processing and training on massive amounts of data, posing high demands on computing power; data is the cornerstone of the application and development of artificial intelligence technology. The large-scale application of artificial intelligence requires training models using massive amounts of data, and without high-quality datasets, there will be no large-scale application of artificial intelligence; algorithms are the direct tools for generating artificial intelligence, and breakthroughs in AI algorithms are the core elements driving the development of institutional artificial intelligence.
AI large models are short for AI pre-trained large models. By pre-training on large-scale data, they can support various applications without the need for extensive fine-tuning. They possess multi-layer neural network structures, advanced optimization algorithms, and powerful computing resources, significantly enhancing the generality and practicality of AI.
"
AI large models possess emergent, scalable, and composite characteristics, which have enabled them to demonstrate superior performance in many aspects.
Firstly, AI large models can lower the threshold for AI development. Traditional AI technologies typically require a large amount of data and computing resources, whereas AI large models can automatically learn and extract features from data through their own learning capabilities, thereby reducing the difficulty of AI development. This enables more developers and enterprises to utilize AI technology to solve problems more easily.
Secondly, large AI models can improve the accuracy and generalization ability of the models. Due to their rich representation capabilities, large AI models can learn more features and patterns during the training process, thereby enhancing the accuracy of the models when handling various tasks.
In addition, AI large models can also apply the knowledge learned on one task to other tasks through methods such as transfer learning, thereby improving the model's generalization ability.
Furthermore, AI large models can improve the quality and efficiency of content generation. In fields such as natural language processing and image recognition, AI large models can generate richer and more accurate content based on input information. This can not only enhance user experience but also reduce operational costs for businesses. For example, in news writing, advertising creativity, and other areas, AI large models can help businesses quickly generate high-quality content.
Finally, AI large models have achieved a breakthrough in traditional AI technology. Traditional AI technology is usually limited to a specific domain or task, whereas AI large models can generalize across multiple domains and tasks through their powerful learning capabilities. This gives AI large models broader application prospects and is expected to promote the widespread application of AI technology in various fields.
In summary, AI large models have achieved a breakthrough in traditional AI technology through their emergent, scalable, and composite characteristics. This has made AI large models valuable in various aspects such as lowering development barriers, improving model accuracy and generalization capabilities, enhancing content generation quality and efficiency, and bringing tremendous development potential to all industries.
The enterprise service large model fully leverages innovative technologies such as knowledge assistants, data insights analysis, and emotional interaction to provide efficient empowerment for four core business scenarios: intelligent marketing, operations, customer service, and digital employees. It enhances the agility and innovation of enterprises in market competition.
"
Enterprise service large models empower intelligent marketing: The enterprise service large model has achieved innovative applications in five directions including customer relationship management, customer profiling insights, content generation, intelligent recommendation, and decision insights. Especially in the field of content generation, the basic capabilities of the large model empower marketing and have become one of the most promising exploration areas. In actual marketing scenarios, this capability effectively meets the needs for text and image generation, greatly improving the creation efficiency of copywriting and posters. By combining high execution with wide use, it provides outstanding support for marketing activities.
Enterprise service large models empower intelligent operations: Intelligent operations transform data into business insights, support intelligent decision-making and service optimization, achieve operational efficiency improvement and customer experience enhancement. Through the collaboration between humans and machines, the best results are achieved, accurately reaching customers, real-time feedbacking market demands, and improving customer satisfaction and competitiveness.
Enterprise service large models empower intelligent customer service: Enterprise service large models have significant advantages in the field of intelligent customer service, capable of automatically answering questions, solving problems, and providing support, effectively reducing the pressure on manual customer service staff.
Enterprise service large models empower digital employees: By leveraging virtual assistants with a high level of knowledge reserve, enterprise service large models optimize interactive experiences in intelligent marketing, operations, and customer service areas. They guide personalized product recommendations, real-time operational monitoring, and automated customer service Q&A, thereby assisting enterprises in achieving more intelligent and efficient operational models.
The enterprise service large model empowers businesses in intelligent marketing, enterprise operations, intelligent customer service, and digital employee services, comprehensively improving business efficiency, customer experience, and operational effectiveness.
"
Intelligent Marketing Business Area: The enterprise service large model empowers customer relationship management, customer data platforms, marketing automation, and marketing content generation in the intelligent marketing field. It supports full-channel traffic diversion, precise targeting of target audiences, automated workflow configuration, and can automatically generate marketing copy to enhance market competitiveness and growth efficiency.
In the field of intelligent operations, enterprise service large models have advantages such as real-time feedback data, human-machine collaborative decision-making, and visualized abstract data in intelligent operations. These advantages help enterprises understand market and customer needs, optimize operational strategies, improve user satisfaction, and enhance brand value.
Intelligent customer service business domain: The enterprise service large model empowers multiple key aspects in the intelligent customer service business domain, including online customer service, voice customer service, intelligent quality inspection, and assisted robots. Through online customer service, it achieves real-time answering to customer questions, automatic handling of services, and provides companion chat conversations. In terms of voice customer service, it can be applied to intelligent bill marketing collection, message notifications, and responses to customer inquiries. In terms of intelligent quality inspection, the large model supports multiple types of customer service product quality testing and can also intelligently detect content compliance. Assisted robots provide excellent support in aspects such as business process navigation, agent script recommendation, real-time backend monitoring, and assisting customer service staff, thereby improving and optimizing customer service efficiency.
Digital Employee Business Area: Enterprise service large models further enhance the service experience of digital employees through multimodal interaction and service guidance. Digital employees can shape a consistent corporate image IP externally and act as outstanding employees internally, improving the interactive experience of knowledge Q&A.
The impact of AI large models varies across industries, achieving significant cost reductions in service-oriented industries but having a smaller effect on manufacturing and basic source industries. In addition, the prospects for AI technology implementation depend on the value space of technical applications and deployment feasibility.
"
AI large models have varying impacts on different industries. Service-oriented and media industries experience cost reductions, while product-oriented and manufacturing industries are less affected. The successful application depends on potential value and feasibility.
The impact of AI large models varies across various industries. In service industries, AI can achieve a significant cost reduction of 5.8%, mainly concentrated in areas such as customer marketing, customer operations, and customer service, which involve acquiring and converting customers. These areas have high substitution potential. In media industries, the cost reduction ratio is 2.8%, mainly reflected in key areas such as sales channel management and marketing content. For product industries, AI is expected to reduce costs by 1.6%. These companies typically invest heavily in product research and development design and marketing, so in the future, they will reshape their work models by automatically generating product models, appearance designs, and promotional materials. However, in manufacturing and basic resource industries, the penetration rate of AI large models is relatively low, with cost reductions of only 0.5%.
The future development prospects of AI large model technology depend on two core elements. Firstly, it is essential to fully recognize the application potential of this technology, that is, to achieve cost savings through efficiency improvements. Secondly, for different industries, it is necessary to consider the feasibility of rapid adoption and deployment of this technology by enterprises.
In addition, the widespread adoption of AI large model technology requires consideration of other key factors. Firstly, the digitalization level of the industry will directly affect the maturity and feasibility of the application of this technology. Secondly, the task fault tolerance rate, that is, the stability of the technology in handling errors or abnormal situations. Lastly, security and compliance requirements will determine the feasibility of the technology in different industries, as some industries have extremely high demands for data privacy and security.
In-depth Insights: With the maturity of large model technology, GPT-4 will ignite a wave of AI development, attracting widespread social attention. In the future, large models are expected to be widely applied and commercialized in various fields, driving innovation and efficiency improvements in enterprise services.
"
From the GPT algorithm proposed by OpenAl with 11.7 billion parameters in 2018, to the launch of GPT-3 with 175 billion parameters in 2020, the number of parameters has increased by 116 times, crossing into the realm of large models in the hundreds of billions, which had a huge impact on NLP models in 2019. Subsequently, OpenAI introduced the GPT-3.5 Turbo model, which has even larger parameters and higher precision, further popularizing artificial intelligence. This year, OpenAl released the multimodal pre-trained large model GPT-4 on March 15th. Its technical principles and training mechanism are similar to those of GPT-3.5, but it has sparked great public enthusiasm for AI. After an upgrade, GPT-4 has enhanced ChatGPT's image recognition capabilities, extended text limits to 25,000 words, improved the accuracy of professional responses, and enhanced its style transformation capabilities. Compared to previous models, one of the most significant innovations of GPT-4 is its multimodal capability.
With the assistance of language large models, vertical applications have shown higher professionalism, high-quality outputs, and outstanding performance in specific tasks. Currently, these applications have widely penetrated into fields such as finance, government affairs, transportation, healthcare, education, etc. The reason why industry large models have achieved success in highly penetrated application fields is that they possess rich and high-quality data, strong technical demands and innovation requirements, as well as a standardized business environment. In addition, the demand for objective and rational advice in these fields also motivates large models to provide more accurate and logical solutions to meet the requirements of advanced decision-making and strategic formulation. This success has not only promoted the development of the industry but also provided new opportunities and impetus for the application of large models.
The impact of AI large models varies across various industries. In service industries, AI can achieve a significant cost reduction of 5.8%, mainly concentrated in areas such as customer marketing, customer operations, and customer service, which involve acquiring and converting customers. It has high substitution potential. In media industries, the cost reduction ratio is 28%, mainly reflected in key areas such as sales channel management and marketing content. For product industries, AI is expected to reduce costs by 1.6%. These companies typically invest heavily in product research and development design and marketing, so they will reshape their work models by automatically generating product models, appearance designs, and promotional materials in the future. However, in manufacturing and basic resource industries, the penetration rate of AI large models is relatively low, with a cost reduction of only 0.5%.
