We develop a model where heterogeneous agents choose whether to engage in on-the-job search (OJS) to improve labor income. The model accounts for untargeted microdata patterns: fiscal incentives affect job-to-job mobility and wage growth of stayers—but not leavers—across the income distribution, pointing to OJS as a key driver of labor costs. Calibrated to micro and macro moments, the model shows that OJS cost shocks significantly affect real activity and inflation. The permanent decline in OJS costs—driven by ICT and AI-based tools—offers a novel explanation for the weakening of the unemployment-inflation relationship documented in empirical studies.