This paper investigates the collaborative practices and computational artifacts that welfare workers use in a public welfare agency. Specifically, the paper focuses on caseworkers’ knowledge practices related to assessing unemployed citizens and identifying ‘perfect’ pathways. I draw upon an ongoing ethnographic study, carried out in one of the largest municipal jobcentres in Denmark. Findings from this research point out that existing computational artifacts support compliance with welfare policy, while limited support is provided to caseworkers in helping citizens obtain an employment. The contribution of the paper is three-folded: 1) identifying fundamental characteristics of the caseworkers’ knowledge work entailed in assessing unemployed citizens and identifying appropriate pathways, 2) examining the conditions surrounding these knowledge practices, and 3) discussing implications for the design of computational artifacts that better support local knowledge practices. While maintaining support to policy compliance, I argue that computational artifacts can also support ‘data-driven knowledge’, meaning the creation of knowledge that is based on data collected from the wide range of cases of unemployed.