大公司，想养“龙虾”也不容易

Core Viewpoint - The article discusses the challenges and opportunities of integrating AI technologies, particularly intelligent agents like "OpenClaw," into non-tech industries such as healthcare and banking, highlighting the need for proper infrastructure, talent, and data management for successful implementation [3][12]. Group 1: AI Integration Challenges - Many large non-tech enterprises, including those in healthcare and banking, are unprepared for AI integration due to a lack of core talent skilled in fine-tuning large models and the inability to utilize internal data effectively [3][12]. - A survey by PwC revealed that 61% of financial institutions have less than 10% of their tech budget allocated to AI, with aspirations to increase this to 50%, but current profit declines limit their ability to invest significantly in AI [6][8]. - The cost of building the necessary infrastructure to support AI technologies is substantial, and companies face budget constraints that hinder their ability to invest in AI [7][8]. Group 2: Talent Shortage - The scarcity of core AI talent is a significant barrier for non-tech companies, as many skilled professionals are concentrated in tech giants like Google and Alibaba, making it difficult for traditional industries to attract them [12][13]. - Existing IT personnel in large enterprises may only be familiar with outdated AI technologies, which complicates the transition to new models that require advanced skills in fine-tuning and reinforcement learning [12][13]. Group 3: Data Management Issues - Data availability and quality are major challenges for AI deployment, particularly in the pharmaceutical industry, where companies struggle to gather sufficient real-world data for effective AI training [13]. - Financial institutions face significant data management challenges related to security and privacy, which limits their ability to leverage internal proprietary data for AI applications [13]. - The lack of standardized data management practices within large organizations can lead to inefficiencies, with data cleaning and processing consuming a significant portion of AI project timelines [13].