This research applies a task-based approach to measure and interpret changes in the employment structure of the 168 largest US cities in the period 1990-2009. As a result of technological change some tasks can be placed at distance, while others require proximity. We construct a measure of task connectivity to investigate which tasks are more likely to require proximity relative to others. Our results suggest that cities with higher shares of connected tasks experienced higher employment growth. This result is robust to a variety of other explanations including industry composition, routinisation, and the complementarity between skills and cities.