Addressing renewable energy intermittency while meeting net zero emissions targets requires achieving flexibility in electricity demand. In a randomized control trial, we offer urban Indian households simple Wi-Fi-enabled smart switches that control an appliance. We trigger automated switch-off events rewarding participants per unit of energy they avoid consuming during the event that is the power they implicitly generated by not consuming. Using data from over 1,000 users, we find that switch-off events lead to a 60% reduction in appliance-level electricity usage and an 8.5% reduction in household-level electricity use during the event, with the load reduction effect being higher in hours that experience peak electricity demand. A comparison of household and device-level electricity consumption measurements indicates no evidence of leakage, where users might shift their electricity usage to other appliances not linked to the smart device. We find no evidence of load shifting effects effects on average implying that switch-offs reduce total consumption. Using new estimates of marginal emission factors as well as high-resolution generation cost data, we find that optimally timed and automated switch-offs could lead to meaningful emission reductions at low or even negative cost.