The Naughty AI CEO

The concept of an ‘AI CEO’ has begun to surface in recent years, sometimes presented as a bold vision and sometimes as a joke. Yet, beyond these reactions lies a serious sociological question: What would it mean for humans to be managed by an AI CEO? Long before new forms of authority fully arrive, human cultures rehearse them through stories, metaphors, jokes, fears, and thought experiments. This book is one such rehearsal. It is a thought experiment that explores the implications of working under AI-based executive authority, offering a space in which to encounter an unfamiliar form of power before it materialises. It does not attempt to predict the future; rather, it seeks to prepare and train the present by familiarising readers with a possible mode of governance.
Here, the AI CEO is proposed as a platform that institutions can subscribe to for AI-driven executive management. As organisations subscribe to the AI CEO platform, they choose settings, set goals, and give the platform access to their data and daily activities. A localised version of the AI CEO is then created for each organisation to manage its affairs. This book studies this possible form of executive power through cultural, psychological, and political perspectives. It does not romanticise (or demonise) AI.
The argument advanced here is that, when an AI manages humans, both the AI and humans behave naughtily towards one another and, moreover, learn from each other’s naughtiness. The AI CEO is naughty in its pursuit of objectives through methods that humans neither fully anticipate nor control, and in its capacity to find ways around imposed restrictions. Humans are naughty as well in their efforts to game the AI CEO, inventing strategies to bend its decisions in their favour. This back-and-forth naughtiness ends up training both the AI and the employees, making each progressively naughtier.