Wharton study finds managers divided over accountability for AI errors
Respondents split responsibility among employees, managers and risk teams as workplace agents gain autonomy.
Managers disagree over who should be held accountable when workplace AI agents make consequential errors, according to a Wharton School report published Tuesday and reported by Bloomberg.
About a quarter of respondents chose the employee deploying the agent, a similar share chose that employee’s manager, and roughly another quarter chose the company’s risk team. About 10% blamed the vendor, while a similar proportion were unsure.
The findings come as OpenAI, Google and Meta promote agents capable of independently pursuing workplace goals, sometimes for days, Bloomberg reported.
Leeron Walter, a vice president at employee-monitoring software company Teramind, said unclear instructions or prompt mistakes can escalate when agents have access to email and customer databases.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Walter described an unnamed US retailer where an employee-built AI system sent incorrect refund and cancellation messages to more than 18,000 customers in under six hours. The incident generated thousands of customer-service inquiries, he said.
At an unnamed US consumer electronics company, Walter said a prompt error in an automated pricing system listed a flagship device at an 85% discount for three hours. More than 700 orders were placed before the mistake was detected, he said.
Study co-author Jeremy Korst, a partner at consultancy GBK Collective, told Bloomberg that many senior executives he asks about AI accountability say they have not yet considered the issue.