In this article, we describe two experiments measuring the impact of a collection of interventions informed by behavioural sciences to reduce unemployment. In a small-scale pilot study (n = 2,383) run in partnership with a Jobcentre in the UK, we found that small changes to the way jobseekers interacted with employment advisers showed promising effects. Based on these findings, we refined our intervention and tested it in a second, larger trial (n = 88,033) across 12 Jobcentres in the UK. We found that our intervention significantly increased off-flow from benefits. These experiments demonstrate that policies and programmes aimed at reducing unemployment can benefit greatly from a deeper understanding of the behaviours of jobseekers and employment advisers. Further, we suggest that this approach could have positive implications for other areas of public policy.