Google DeepMind’s economic analysis treats an AI’s ability to solve a task and multiple agents’ ability to coordinate as separate dimensions. It argues that even gradual improvements in individual competence can have nonlinear economic effects once coordination crosses a threshold enabling larger organizations and bundles of work. Standard benchmarks focus mainly on individual tasks, leaving complementarities and coordination failures undermeasured. This is a theoretical framework rather than an empirically established employment forecast; its practical implication is to evaluate role coordination and shared outcomes alongside individual agent performance.