We evaluate an experiment in which randomly encouraged households activate a smart-thermostat feature that automates responsiveness to time-of-use electricity pricing. The thermostat feature reduces electricity use by raising indoor temperatures, thus increasing thermal discomfort in some households during peak periods. Changes in discomfort are small, concentrated among households who spend the most time at home, and do not prompt them to adjust the feature’s intensity or deactivate it. Using energy cost savings and experienced indoor temperatures, we calculate households’ revealed preference trade-off between comfort and cooling expenditure and find that households are willing to trade off small monetary savings for small increases in discomfort. Automation thus provides a low-cost opportunity to make small changes in energy demand at the household level, with potentially large electricity supply-cost reductions at scale.