In 2025, AI was caught in the "generative AI paradox" where, despite its high adoption rate among companies, the percentage of people who realized the benefits of its adoption was low . VCs investing huge amounts in the AI field are under pressure from their fund LPs to deliver returns, and a sense of crisis about the AI bubble is beginning to spread.
Under pressure from investors, AI assistants are billing themselves as AI agents, leading to widespread "agent washing," which disappoints companies that have adopted them. Gartner 's published evaluation model for the suitability of AI agents is emblematic of the current state of confusion in the AI world.
AI agents perform tasks autonomously, but these are ultimately individual skills. To use a soccer analogy, an ace striker alone cannot win a game. Team play is achieved with a commanding back and a guardian goalkeeper. Furthermore, the team can achieve its goals with the help of a manager.
This trend has inevitably led to the development of multi-agent systems that fight with organizational strength, and further to the development of agent-type AI with a supervisor, which is moving towards a more complete system of organizational strength. Many startups that provide development frameworks are also launching products that realize multi-agent systems one after another.
While AI agents are capable of persistently repeating loops of inference and execution toward their goals, they also incorporate mechanisms for human intervention at appropriate times. An increasing number of frameworks are available that allow for the development of such highly functional mechanisms using no-code or low-code. With the barriers to implementation lowered, concrete benefits of agent-based AI are expected to emerge in 2026 .
Report format: PDF (7.9MB)
Original data: PowerPoint, 78 slides, A4 size