Artificial general intelligence (AGI) refers to a hypothetical AI system with broad, flexible competence - one that can learn and perform across the full range of tasks a person can, rather than being narrowly skilled at one job. It is contrasted with the “narrow” AI that defines essentially all systems in use today, each of which is good at a specific function (translating, recommending, generating images) but cannot transfer that skill freely to unrelated problems.
The term was popularized by the 2007 volume “Artificial General Intelligence” edited by Ben Goertzel and Cassio Pennachin, published by Springer, which gathered researchers around the explicit goal of building broadly capable systems rather than task-specific ones. Crucially, there is no single agreed definition. Different labs set their own bars: OpenAI, for instance, has publicly framed AGI as highly autonomous systems that outperform humans at most economically valuable work. Because each definition draws the line differently, “AGI” functions as much as a mission statement as a technical specification.
AGI is also a moving target. Capabilities once treated as hallmarks of general intelligence - playing chess, holding a conversation, passing professional exams - have been achieved by narrow systems without anyone agreeing that AGI arrived. This pattern, sometimes called the “AI effect,” means the goalposts shift as each milestone falls, which is part of why the term resists precise measurement.
Why business readers should care: AGI is a load-bearing word in AI strategy, fundraising, and policy debate, yet it means different things to different speakers. When a company invokes AGI, the useful question is always: by whose definition, and measured how? Treat specific, testable capabilities as the real signal and the AGI label as a claim to be unpacked.