Scholarlink

<p>Artificial intelligence plays a critical role in automating jobs. Early application of artificial intelligence primarily relies on expert systems to automate routine jobs that follow predetermined rules. Recent advancement in machine learning (ML) has ushered in a new and fundamentally different mechanism for job automation, in which rules are derived by ML models from data and then applied to automation. In this research, we extend existing studies on job automation and argue for the necessity of explicitly incorporating both mechanisms of job automation: task routineness and ML model appropriateness. We demonstrate that expert systems are mainly used for automating routine jobs, whereas model appropriateness is more effective for automating nonroutine jobs. Furthermore, the consequence of error exhibits an inverted-U shaped relationship with job automation, and this relationship varies across different types of jobs. Our study sheds new light on the automation-augmentation paradox and offers both theoretical insights and practical guidance for managers making AI-enabled automation decisions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Artificial intelligence and job automation: opening the “black-box” of the underlying mechanisms

  • 期刊论文

  • 发表日期：

    2026年4月11日

查看全文
  • Gang Peng,
  • Yi Jenny Zhang,
  • Yung-Ming Li

摘要

Artificial intelligence plays a critical role in automating jobs. Early application of artificial intelligence primarily relies on expert systems to automate routine jobs that follow predetermined rules. Recent advancement in machine learning (ML) has ushered in a new and fundamentally different mechanism for job automation, in which rules are derived by ML models from data and then applied to automation. In this research, we extend existing studies on job automation and argue for the necessity of explicitly incorporating both mechanisms of job automation: task routineness and ML model appropriateness. We demonstrate that expert systems are mainly used for automating routine jobs, whereas model appropriateness is more effective for automating nonroutine jobs. Furthermore, the consequence of error exhibits an inverted-U shaped relationship with job automation, and this relationship varies across different types of jobs. Our study sheds new light on the automation-augmentation paradox and offers both theoretical insights and practical guidance for managers making AI-enabled automation decisions.