Caseworkers are the primary human resource used by governments to match unemployed workers to jobs. This paper provides first estimates on the role of caseworkers in reducing the duration of unemployment spells, using register data from the Swiss UI. For identification, I exploit caseworker-time specific variation in unexpected work absences, which I show to be independent of job seeker characteristics and labor market conditions. I find that the duration of unemployment increases on average by 2 weeks if a job seeker's caseworker is absent for at least ten workdays during the first three months of unemployment. This effect is largely driven by absences of high value added caseworkers. As main channels, I identify a reduction in the number of caseworker meetings and an increase in the probability of meeting with a replacement. Further, the participation in programs with individual-specific content is reduced. It appears that personal interactions and individualized treatment assignments are an important component of the caseworker's production function. Placebo tests reveal that future absences do not affect current outcomes, confirming the absence of confounding caseworker-time specific trends.