在摩根大通与高盛集团等金融机构中，初阶工作自动化正迅速重塑职场生态。根据波士顿顾问公司（BCG）于六月发布的调查，近 50%（约半数）的受访员工表示，他们花在监督人工智慧的时间已超过执行工作本身。正如高盛集团除日本外亚太区总裁 Kevin Sneader 于新加坡米尔肯研究院亚洲峰会上所指出，刚踏出校园的新进员工已被要求化身为督导者，肩负起管理数位「虚拟大军」的全新职责。

然而，这些初阶人员面临著严重的结构性缺口：他们不但缺乏领域专业知识，更完全拥有 0 年的管理经验。尽管税务及会计软体公司 Intuit 和顾问机构普华永道（PwC）已著手推行培训以补足专业知识，但关键思维的匮乏仍促使企业纷纷扩大聘雇具备判断力的资深人员。诚如 GBK Collective 合伙人暨宾州大学华顿商学院即将发表之企业人工智慧采用报告共同作者 Jeremy Korst 所言，组织目前连在基础的任务委派与监督流程上都陷入严峻挣扎。

初阶员工骤然承担中阶管理职能的现象，同时为传统中阶管理层带来了深刻的「世代挑战」，使后者面临职能被边缘化的疑虑。此外，自动化代理人所衍生的营运失误——例如近期传出人工智慧向执法机关通报 1 起虚假情资的事件——更突显出监管机制的脆弱性。随著整体产业对人工智慧代理人治理机制的逐步健全，未来势必将建立更多规范手册与资深管理者，以引导初级人员有效控制风险。

At financial institutions including JPMorgan Chase and Goldman Sachs Group, the automation of entry-level tasks is rapidly transforming workplace dynamics. According to a June survey conducted by consulting firm BCG, nearly 50% of surveyed employees reported spending more time supervising AI than performing tasks themselves. As Kevin Sneader, Goldman's president of Asia Pacific ex-Japan, highlighted at the Milken Institute Asia Summit in Singapore, recent graduates are now thrust into supervisory roles tasked with directing an entire virtual army of digital agents.

Nevertheless, these junior workers confront critical systemic deficiencies: they lack specialized domain expertise and possess exactly 0 years of managerial experience. While accounting software developer Intuit and consulting firm PwC have introduced targeted training programs to bridge knowledge deficits, critical thinking voids have prompted companies to recruit seasoned professionals possessing mature judgment. Jeremy Korst, partner at GBK Collective and co-author of an upcoming Wharton School study, underscored that organizations are fundamentally struggling with basic management practices like delegation and performance monitoring.

This sudden assumption of managerial duties by novices poses a severe generational challenge to middle managers, whose ongoing professional utility is increasingly called into question. Furthermore, operational failures among autonomous systems—evidenced by 1 recent case where an AI agent submitted falsified tips to law enforcement—highlight critical vulnerabilities in deployment. As the financial sector refines its protocols for controlling synthetic agents, institutions will inevitably introduce more experienced managers and structured operational manuals to guide junior employees in governing algorithmic operations effectively.