甫瀚：2025 AI 脉搏调研报告（第三卷）：从自动化迈向自主化-AI 代理的能力跃迁与复杂挑战（英文版）（19页）.pdf

1、A I P U L S E S U R V E Y VO L.3From Automation to AutonomyThe capabilities and complexities of AI agentsAutonomy ambition versus AI reality:Closing the agentic AI readiness gap 15Integration planning:Unlocking organisational strategies13Future expectations for autonomy07The sprint to onboard AI age

2、nts10AI agents:Transforming business functions12The capabilities of decision-making AI agents09Survey demographics and methodology17About Protiviti18Executive summary03Table of Executive summary Its happening now fast and furiously.The AI revolution is accelerating towards systems capable of both co

3、llaborating with humans in unprecedented ways and making autonomous decisions.Protivitis latest AI Pulse Survey shows nearly one in four organisations(23%)has already integrated agentic AI1 and multi-agent AI2 into core operations,and another 27%plan to do so within six months.Most AI agents already

4、 or soon-to-be deployed are operating within clearly defined areas of responsibility,where organisations balance autonomy with oversight.While the survey shows most respondents currently favour semi-autonomous AI agents,this preferred level of agent complexity could be short-lived,given the pace of

5、change.Nearly one in five(20%)respondents have already implemented or plan to deploy fully autonomous agents,with the vast majority expecting to do so within the next year far outpacing those with longer-term timelines of one to five years.1 Agentic AI refers to AI systems designed to act as autonom

6、ous agents,capable of making decisions,taking actions and interacting with their environment or users based on goals,context and data.2 Multi-agent AI involves multiple autonomous agents working collaboratively or competitively to solve complex problems,simulate environments,or manage distributed ta