Within three years, 80 % of companies had adopted generative AI , but the paradox is that 80 % have not realized any benefits. This situation will change from the second half of 2025 , and with the spread of AI agents, the benefits are beginning to be felt as they perform tasks that people would normally have to do themselves.
The first use case was e-commerce , where traffic via AI agents is rapidly increasing. Furthermore, we are beginning to see a trend where leads from AI agents have higher conversion rates than leads from humans. Therefore, the key question is how many products and services AI agents are secretly connecting with.
The traditional method of connecting AI agents to external tools using APIs allows for the combination of up - to-date information and tools that have not been learned in advance. However, the inefficiency of development caused by individually responding to the specifications and constraints of external tools created a bottleneck in scalability. Anthropic solved this problem with MCP , released in 2024 , and after opening it up in 2025 , it is becoming the de facto standard in the industry.
Learning from the lessons of APIs , the more a connection layer becomes a commodity, the less differentiated it becomes, and in the horizontal domain, where versatility is emphasized, only cost-effectiveness is emphasized. In contrast, in the vertical domain, complex expertise and problem-solving give rise to advantages that cannot be easily imitated, creating high barriers to entry and ensuring the sustainability of the business.
Fragmentation creates a market for standardization, but history has shown that the connectivity layer as a function has no value. This requires a clear survival strategy, such as targeting areas where originality and differentiation can be achieved, or shifting to value performance rather than cost.
Report format: PDF (6.3MB)
Original data: PowerPoint, 78 slides, A4 size