What percent of tasks can AI automate?

No. It is a directional mnemonic. The real finding is that work always splits into three buckets, automate, augment, and human, with meaningful weight in each. The exact sizes shift with the function. Knowledge work sits closer to a third automatable, frontline work much less. The claim is not that every role is exactly 30/40/30. It is that three buckets always emerge.

Because companies have not deployed AI to the ceiling yet. Across 2,400+ companies, the average role today sits near 9% automated and 23% human-only. The roughly 30% figure is what becomes possible when an enterprise adopts current-generation agentic AI in full. The distance between 9% and the ceiling is the operational work of the next several years.

Anthropic's Economic Index analyzed roughly 4 million Claude conversations and reported 52% augmentation, 45% automation. That is a two-bucket model with no human-only category, because everything in the sample already had AI in the loop. The 30/40/30 pattern adds the third bucket so enterprises can plan around the work that must stay human. It extends the Anthropic finding rather than contradicting it.

Roles where almost every task is language, data, or pattern work. At the enterprise ceiling, Market Research Analysts sit near 85% automate and Accountants near 62%, because the core tasks can run end to end. Roles dominated by physical work, patient contact, or accountability stay augment-heavy. The only way to know a specific role is to classify its actual tasks.

Not necessarily. A high automate share means the routine tasks compress and the role shifts toward supervision, exception handling, and judgment. The design goal is that people get better at their jobs because low-value tasks get automated and their time moves to the work they were hired for. The percentage describes the tasks, not the headcount.

Classify the tasks. Paste a job description into the analyzer and it returns the automate, augment, and human-only split for that specific role in under a minute. For a department or workflow, the same task-level classification rolls up into a percentage you can plan against, rather than a generic industry average.