AI is in 68 percent of jobs, but only 21 percent of the tasks

Google measured real Gemini usage across the economy. The breadth is striking: sixty-eight percent of occupations show AI use, spanning ninety percent of employment, from lawyers to mechanics. The depth is not. Within any given job, AI touches only about twenty-one percent of tasks, and fewer than ten percent of interactions fully automate a task. It is mostly ideation, drafting, retrieval, and learning.

And most of it is not at work at all. Eighty-six percent of AI interactions happen outside the job, on personal admin, research, and learning. People are getting fluent on their own time.

Nearly everyone has touched AI. Almost no one has rebuilt their work around it. That gap, between trying it and working in it, is where the actual productivity still sits untouched. The people and teams that go deep, past the easy fifth of tasks, are effectively working with a different tool than everyone else.

Breadth is done. Depth is the opportunity. The question is not whether your team uses AI, it is how much of the work it has actually changed.

People already build the habit in their personal life. The workplace gap is about carrying it into real tasks, not teaching from zero.

Augmentation, not automation, is the honest near-term picture. Plan for people plus AI, not people replaced.

Audit one role for the eighty percent of tasks AI does not yet touch. That list is your roadmap.

Copy what your best AI users already do off the clock. The methods exist, they are just not shared.

Go deep on one workflow before going wide across ten. Depth compounds, breadth plateaus.

The shallow-usage number will rise. Today's twenty-one percent is a moving floor, not a ceiling.

Personal-account usage carries data and security exposure. Convenience is outrunning governance.