This study explores the impact of recommendation systems’ algorithms on individuals’ teammate choices when assembling teams. I posit that different algorithmic designs will lead individuals to choose one team member over another, thereby shaping their final teams’ composition, dynamics, and performance. To investigate this, I conducted a between-subject laboratory experiment (N=332 participants) using four team formation algorithms, each manipulating participants’ agency and the inclusion of diversity criteria. The findings show that the algorithms directed participants toward different team combinations, resulting in varied communication patterns and team viability levels. Notably, algorithms that implemented either diversity criteria or agency did not result in high team performance. While allowing personal agency without any diversity criteria led to homogenous teams with low performance, incorporating diversity criteria alone resulted in teams with low communication frequency and viability. In contrast, the team formation algorithm that combined personal agency with diversity criteria facilitated the assembly of effective diverse teams. I conclude by discussing the findings and the implications of algorithmic designs on organizational capabilities.