AGI vs narrow AI isn’t the real question anymore

A Yale economist recently suggested AGI might not automate most jobs, not because it can’t, but because it may not be worth the effort. It’s a small shift in framing, but quite a revealing one.

Most of the AI people use today is still what’s known as narrow AI: systems designed for specific tasks like writing, analysing or generating content. AGI, or artificial general intelligence, refers to the idea of a more flexible system that can apply knowledge across different situations, adapt independently, and operate more like a human thinker.

The assumption is often that once AI can do something, it inevitably will. But work isn’t just a checklist of tasks waiting to be automated.

Some jobs are messy, relational or unpredictable in ways that are difficult to systemise. Others simply aren’t valuable enough to justify the cost and complexity of replacement, even if the technology exists. Which makes parts of the AGI debate feel slightly ahead of themselves.

Because right now, it’s narrow AI quietly changing how work gets done: speeding things up, shifting expectations, and altering where human judgement sits in the process.

Maybe the more useful question isn’t when AGI arrives. It’s what we’re choosing to automate in the meantime, and why.