Unaccountable Delegation, Fading Skills: Mapping the Risks of Workplace AI Agents
Gabriele La Malfa, Lakmal Meegahapola, Edyta Bogucka, Jie M. Zhang, Michael Luck, Elizabeth Black, Daniele Quercia
Why It Matters
What makes this one worth your time
Understanding specific AI risks in the workplace is crucial for designing safer human-AI collaborations and mitigating potential skill erosion and oversight issues.
The paper introduces a framework and taxonomy for identifying and classifying job-specific AI risks in the workplace.
Summary
The paper develops a multi-layer framework to identify job-specific risks of AI agents in the workplace, applies it to job tasks from the O*NET database, and validates the resulting risk scenarios with workers and an LLM judge. It extends an existing taxonomy to classify these risks into 15 categories, highlighting the nuanced risks of AI augmentation and automation.
Key contributions
- Development of a multi-layer framework for modeling AI agent risks in the workplace.
- Creation of a 15-category taxonomy for classifying workplace AI agent risks.
- Validation of risk scenarios with workers and an independent LLM judge.
Notable insights
- Augmentation can lead to skill erosion and oversight issues, challenging the assumption that it is inherently safer than automation.
- Erroneous Agent Actions at the human-agent boundary present significant risks, emphasizing the need for careful design of human-AI interactions.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2608.08601v2 Announce Type: replace Abstract: To anticipate socio-technical risks from AI agents, organizations need taxonomies to classify them. However, existing AI risk taxonomies focus on broad risks and do not capture job-specific risks introduced by agents. To address this gap, we make three main contributions. First, we developed a multi-layer framework from a literature review of AI agents. The framework models three core components and their interactions: agents, goals, and environment. Second, we embedded this framework in a structured prompt and applied it to descriptions of 2,078 job tasks from the O*NET database, producing 8,356 risk scenarios labeled by severity and deployment mode (automation or augmentation). We validated these scenarios with 45 workers across 10 job roles and an independent LLM judge, confirming their plausibility and alignment with job tasks. Finally, we extended an existing taxonomy to create a 15-category taxonomy of workplace AI agent risks that covers all our risk scenarios. Our analysis highlights four findings. First, augmentation is not inherently safe because overreliance on agents can gradually erode workers' skills and oversight. Second, Erroneous Agent Actions accounts for the largest share of risk scenarios and has the highest concentration of severe risks. Many arise at the human-agent boundary. Third, automation is associated mainly with organizational risks, while augmentation is associated mainly with risks to workers. Fourth, workers found our taxonomy easier to use for a risk classification task than two other taxonomies and preferred it in 64% of non-tied comparisons with a recent generative AI risk taxonomy. These findings show that workplace AI agent risks do not arise from agents alone; they also depend on how people work with agents and how agents are deployed. Safer workplaces require not only safer agents but also carefully designed human-AI agent collaboration.