Important Notice: This material does not constitute any form of offer, commitment or other legal document from Freshwater Capital, nor is it professional advice in any investment, legal or financial aspect. Past performance is not indicative of future results. Investment involves risks. Please exercise caution.
The debut of DeepSeek R1 triggered a global reevaluation of Chinese technology assets. Since then, a plethora of “AI assistants,” with Mauns as a prime example, have emerged, signifying AI’s deepening integration into production and daily life. Unlike the previous internet wave, AI applications exhibit unique development characteristics.
During the internet and mobile internet eras, the core value of scenarios and platforms was to accelerate information dissemination by enhancing matching efficiency and reducing transaction costs. The rapid spread of information essentially represented a transformation in production relations, such as distribution methods, which led to the relatively fast development of internet applications. In contrast, AI applications, as value – creation tools, primarily boost productivity, optimize operational processes, and improve efficiency, gradually altering user habits. Their development is centered around the productization process based on large – model capabilities and is characterized by slow penetration.
Two Main Categories in the Early Development Stage
Currently, AI applications are still in their infancy and can be technically classified into two categories.
“AI+” Applications: Function Expansion
The first category is “AI+” applications, which integrate AI capabilities into existing platforms or applications to expand and enhance their functions. Examples include office software, search engines, and local – life – related software. For example, Microsoft 365 Copilot allows users to generate documents and analyze data using natural language commands. This category also encompasses some fast – growing but limited – potential efficiency tool applications, like those that automatically generate meeting minutes from voice recordings.
Native AI Applications: New Scenario Creation
Native AI applications form the second category, where AI capabilities are the core product elements. During development, they fully leverage AI functions to create novel application scenarios and user experiences. Examples are chat tool platforms such as DouBao and Kimi, developed by large – model manufacturers, as well as related products in the comics, games, and music fields. Although the category is not yet fully developed, it has high requirements for model capabilities.
Mismatch between Popularity and Usage
Despite the wide variety of AI applications, the “2024 First Half AIGC APP Traffic and Scenario Research Report” by QuestMobile reveals that over 40% of AI application software experienced a decline in traffic in the past year. There is a mismatch between popularity and usage, indicating a discrepancy between users’ scenario – based demands and the capabilities offered by these apps. The lack of user stickiness in products is mainly because relying solely on technology – driven enhancements of large – model capabilities is insufficient to build a solid product moat, especially as high – performance large models become more accessible, further narrowing the technological gap.
The Key to Commercial Breakthrough: Meeting Scenario – based Demands
The key for AI applications to achieve commercial breakthroughs lies in meeting users’ scenario – based demands. ChatGPT serves as a prime example. In its early days, it faced a slowdown in user activity growth. However, OpenAI reversed the situation through continuous iterations, launching more powerful models and features. For example, in April – May 2024, the release of GPT – 4o brought multi – modal capabilities, supporting real – time voice interaction and image recognition, enabling users to get answers to math problems by simply taking a photo. In July – August 2024, the Advanced Voice Mode was introduced, significantly improving conversation fluency.
When evaluating the product strength of AI applications, technical elements alone are not enough. It is also crucial to consider factors such as continuous financial investment and the accumulation of proprietary data, including operation and user data in specific industry fields, to support intelligent exploration. These factors jointly determine whether a product can succeed in the highly competitive market, gain user favor, and achieve commercial value.
Commercial Scenarios: B – end and C – end with Different Focus
From the current development of AI applications, commercial scenarios can be broadly divided into two categories: B – end and C – end, each with its own development focus.
B – end: Faster Implementation and Measurable Value
In B – end scenarios, the implementation is relatively quick, and the commercial value is more measurable. Even for relatively simple closed – loop business models, customers are often willing to pay, as these models can bring direct transformation results, such as cost reduction, efficiency improvement, new revenue sources, and brand value enhancement.
B – end AI applications develop gradually in a scenario – specific and phased manner. In the early stage, they are mainly applied in simple scenarios or integrated into existing processes and software to handle repetitive tasks with high fault tolerance. These applications have received positive market feedback and generated some user willingness to pay. Currently, B – end AI applications are making rapid progress in general efficiency tools and vertical fields, such as knowledge – based work with high human – labor content and strong repetitiveness, like code programming, text, and data processing, as well as job positions with complex work processes but weak business binding and independent working environments, such as sales, customer service, and HR.
In the medium and long term, B – end AI applications will evolve towards building industry – specific AI applications in vertical fields or developing general and generalized large models and intelligent agents to address more complex B – end intelligent scenarios. In the short term, the application scenarios will still mainly focus on transforming existing business processes. For example, some overseas SaaS software enterprises optimize the customer experience by integrating AI technology into their original business processes.
Source: Salesforce official website, CSDN application collection
C – end: Commercialization Challenges and Leading – Enterprise Advantages
The core of C – end applications is to meet individual immediate needs and drive changes in user behavior. This means that different demand scenarios require separate considerations, resulting in a relatively slow commercialization process.
In light – decision – making scenarios, such as watching short videos or movies, users are more focused on the exploration and experience process, and the need for AI intervention is relatively low, with limited transformation value. However, in heavy – decision – making and result – oriented scenarios, such as search, navigation, life services, text processing, and content generation, or in scenarios with strong personalized content demands, such as social media, games, and education, AI has great potential. For example, in result – oriented scenarios, when a user searches for a travel destination, AI can provide personalized travel guides, hotel reservation services, and other comprehensive information based on the user’s preferences and needs. In content – demand scenarios, AI technology can enable more intelligent NPC behaviors and dialogues in the gaming field and offer more personalized language learning experiences and course recommendations in the education field.
Overall, C – end applications require AI vendors to have a profound understanding of market trends, customer demands, and scene characteristics. Given the diversity and variability of C – end scenarios, it is difficult to apply standardized processes, so higher large – model technical capabilities are needed as support.
In the medium to long term, the “killer app” products for C – end AI applications are more likely to come from leading internet companies. Currently, major internet companies may be slightly behind specialized leading large – model vendors in technological exploration, but the gap is not significant. When the development of large – model technology reaches a plateau, technology will no longer be the core factor limiting product performance. At that time, major internet companies are expected to build a late – mover competitive advantage with their strong financial strength, massive data resources, mature ecosystem, and powerful platform capabilities, and play a more important role in subsequent application development. Looking ahead, we believe that the development of C – end applications will still be largely led by leading internet companies. Currently, we can observe that the capital expenditures of leading companies in the AI field continue to increase, and they are conducting in – depth product exploration in the AI application field.
Related topics：
