•Breaking the Paradox
–Generative AI has been introduced by 80% of companies in three years, but the paradox is that 80% have not realized the effect. The situation will also change from the second half of 2025, and the spread of AI agents is beginning to show the effect of having people do what they should do on their own.
•
•EC via agent
–The first use case is e-commerce, where traffic via AI agents has skyrocketed. Not only that, but AI agent leads are starting to convert more than human leads. The important question is how many products and services the AI agent can connect with behind the scenes.
•
•API to MCP
–The traditional method of connecting AI agents with external tools realizes the combination of the latest information and tools that have not been learned in advance. However, due to the individual response to the specifications and constraints of external tools, development inefficiency was a bottleneck in scalability. The solution is Anthropic's MCP in 2024, which is becoming the de facto of the industry after opening up in 2025.
•
•Winners and losers
–From the lessons learned from APIs, these connection layers become less differentiated as they become commoditized, and only cost-effectiveness is emphasized, especially in the horizontal area where versatility is emphasized. In contrast, in the vertical field, by solving complex expertise and issues, you gain an advantage that cannot be easily imitated, which becomes a high barrier to entry and ensures the sustainability of the business.
•
•The Future of MCP
–Fragmentation creates a market for standardization, but history shows that the connectivity layer as a function has no value, and a clear survival strategy is required, such as targeting areas where uniqueness and differentiation can be acquired, or shifting to performance values that are different from cost