Automation rate
The share of work an AI system completes without a person doing it. A useful headline number that says nothing on its own about the human attention left behind.
Automation rate is the proportion of tasks, steps or cases handled by an automated system rather than by a person. It is the most common way to describe how much an AI deployment has taken over, and it is easy to compute.
The trouble is what it leaves out. Automation can remove execution and leave the work: approvals, context supply, review, correction and exception handling can all stay with people, or even grow. A system with a high automation rate may still demand a lot of human attention per result, because the rate counts who did the visible task, not what it took to get a usable outcome.
Pair it with something honest
Measure supervision load next to it: how much attention does one dependable outcome require, and does that grow with volume? A rising automation rate with flat supervision load is genuine leverage. A rising rate with rising supervision is production speed without operating leverage.
Be explicit, too, about whether the person is staying in the loop or has become the loop. The metric should be re-baselined when the runtime changes.
Read more in The hidden metric in AI automation is supervision.
Related terms
Supervision load
The human attention an AI system still consumes — context supply, review, approval, correction and exception handling — measured per dependable, completed unit of work.
Dependable outcome
A result from an AI-assisted workflow that is actually usable without further repair — the right denominator for measuring the human attention and cost AI automation really consumes.
Human in the loop
A design where a person reviews, approves or corrects an AI system's work at defined points. In practice it can mean a person with real judgment, or a person who has become the loop.
Used in these essays
OpenAI Dots is a test of whether AI can carry a goal, not just complete a task
Dots matters less as another capable assistant than as a test of persistent delegation: can AI keep carrying a goal without giving the user a new system to manage?
The hidden metric in AI automation is supervision
As AI moves from assisting work to leading it, hours saved stop telling the whole story. The scarce resource shifts to human supervision: approvals, exceptions, context and judgment.
Your chatbot is borrowing from the next interaction
AI customer service is usually measured one interaction at a time. But a failed automated interaction can change which channel a customer chooses next time. That makes future adoption part of the economics, not a separate trust metric.