This project builds on the lessons from Robodebt to address the need to understand the working, and impact, of automated processes in government. The problem with robo-debt lay not in the idea, but in the ¬execution. The system was riddled with design flaws. Robodebt is a textbook example of how not to deploy technology in government decision-making. An initiative designed to save revenue has instead led the federal government to repay $721m to 373,000 people. However, care must be taken not to draw the wrong lessons¬. One flawed program does not destroy the case for governments to use technology to improve services and ensure public money goes to the right recipients.