AI agents raise new questions about accountability at work

AI agents are emerging as a new kind of workforce in companies, but organization, accountability and governance have not kept pace, according to AI expert Joel Sunnehall. In his new book, När AI blir en medarbetare, he argues that companies must decide how digital employees are organized and controlled.

In an interview with Techtidningen, Sunnehall says the book is aimed at CEOs, senior leaders and CIOs in Nordic corporations. It examines five areas that need to change as AI becomes part of the workforce: leadership, governance, work, skills and the ability to manage change. The book is based on surveys, interviews and operational experience, and includes a self-assessment test for management teams.

An EY Nordic survey from 2025 found that 26 percent of Nordic CEOs were actively involved in shaping strategies for new technology such as AI, compared with 49 percent globally. The same survey found that 53 percent of companies struggle to assign clear responsibility for new AI technology.

Sunnehall proposes an “agent contract” defining an agent’s mandate, access, triggers and handover points, with a named person responsible. He also says agents should be identifiable and their actions traceable. For decisions with major consequences, such as recruitment and large credit decisions, he recommends that AI prepare the material while a human makes the decision.