智能体是新宠，但非万能药——专访麦肯锡全球资深董事合伙人周宁人

Core Viewpoint - The article discusses the current state and future potential of AI deployment in various industries, particularly in finance, highlighting the challenges and opportunities associated with scaling AI applications effectively [5][14]. Group 1: AI Deployment Status - Despite 88% of global enterprises using AI in at least one business function, only 39% report profitability from these applications, indicating a significant gap between adoption and effective implementation [5][16]. - In China, 83% of enterprises regularly use generative AI in at least one function, surpassing the global average, with 45% achieving large-scale or comprehensive deployment, higher than the global average of 38% [8][9]. - The financial sector is seen as a leading area for AI application, yet many institutions report that the effectiveness of AI does not meet expectations, primarily due to the exploratory phase of AI integration [14][16]. Group 2: AI Agent and Agentic AI - AI Agents, which enable machines to take action, are becoming popular, with 62% of organizations experimenting with them, but less than 10% have fully integrated them into business processes [10][11]. - The emergence of Agentic AI, which allows AI to autonomously complete tasks, is identified as a key trend, driven by reduced inference costs and the rise of smaller models that can operate on local devices [12][14]. - Successful AI organizations tend to have clear AI roadmaps and actively integrate AI into their core processes, moving beyond mere experimentation [14][15]. Group 3: Challenges and Recommendations - Organizations must redesign workflows to effectively leverage AI, focusing on identifying core pain points to enhance collaboration between AI and human workers [16][17]. - It is crucial to avoid a one-size-fits-all approach to AI deployment; instead, organizations should tailor AI applications to specific business needs and ensure proper governance and monitoring [18][20]. - The financial sector must balance innovation with risk management, employing a mixed strategy of rule-based AI for predictable tasks and AI-assisted processes for more complex scenarios [18].