• Returns on AI Investments
– In 2025, AI has fallen into the "generative AI paradox" in which the percentage of people who feel the effects of introduction is low, despite the high penetration rate of companies. VCs who invest huge amounts in the AI field are under pressure to return from the fund's LPs, and a sense of crisis about the AI bubble is beginning to spread.
• Agent Washing
– Due to such investor pressure, "agent washing" is rampant, in which AI assistants claim to be AI agents and betray the expectations of the adopting companies that accept them. The evaluation model for the eligibility of AI agents released by Gartner symbolizes the confusion in this AI community.
• From individual skills to team play
– AI agents perform tasks autonomously, but only individually. If you compare it to soccer, you can't win a game with just an ace striker. There are Bugs in the command tower and Keeper, the guardian deity, so team play is established. In addition, the team can achieve its goals by having the coach take charge.
• From single to multi
– From this trend, it inevitably develops into a multi-agent system that fights with organizational power, and furthermore, it is moving towards a complete system as an organizational power with an agent-type AI with a supervisor. Many startups that provide frameworks for development have also launched products that realize multi-agents.
• Inference & No-Code
– AI agents can persistently repeat the reasoning and execution loop towards a goal, while also having a mechanism to intervene with humans at the right time. There are an increasing number of frameworks that can develop such high-performance mechanisms with no-code or low-code. Since the hurdles for introduction have been lowered in this way, it is expected that the introduction effect of agent-type AI will be concrete in 2026.