A preregistered cross-national study released an open index that measures how much of each occupation agents can take over, broken down by task and country. The research defines four levels of delegation. At level one, a human does the work alone. At level four, an agent completes the work with no human involvement. The index applies that scale across hundreds of occupations. The result is a public dataset and a queryable server that product teams can use to check where agent involvement is plausible. The study shifts the question from whether AI touches a job to how much of a task a human can hand over and under what conditions. That framing maps onto design decisions about when to show a confirmation step, when to let an agent run unsupervised, and when to surface a result for human review. The four-level scale gives product teams a shared vocabulary for scoping what an agent should attempt before a human reviews the output. The index varies by country, so the same task may carry a different delegation level depending on regulatory, cultural, or infrastructure conditions. A designer building for multiple markets cannot assume a single handoff point. Teams can query the open index without replicating the research.