Engineering of nonribosomal peptide synthetases (NRPSs) has faced numerous obstacles despite being an attractive path toward bioactive molecules. Specificity filters in the nonribosomal peptide assembly line determine engineering success, but the relative contribution of the adenylation (A) and condensation (C) domains is under debate. In the engineered, bimodular NRPS sdV-GrsA/GrsB1, the first module is a subdomain-swapped chimera showing substrate promiscuity. On sdV-GrsA and evolved mutants, we have employed kinetic modeling to investigate product specificity under substrate competition. Our model contains one step in which the A-domain acylates the thiolation (T) domain and one condensation step in which deacylation of the T-domain occurs. The simplified model agrees well with the experimentally determined acylation preferences and shows that the condensation specificity is mismatched with the engineered acylation specificity. Our model predicts that product specificity changes during the course of the reaction due to dynamic T-domain loading and that the A-domain overrules the C-domain specificity when the T-domain loading reaches a steady state. Thus, we have established a tool for investigating the poorly accessible C-domain specificity through nonlinear kinetic modeling and gained critical insights into how the interplay of the A- and C-domains determines the product specificity of NRPSs.