Non-ribosomal peptide synthetase (NRPS) enzymes form modular assembly-lines, wherein each module typically governs the incorporation of a specific amino acid or related monomer into a short peptide product. Modules are comprised of one or more key domains, including adenylation (A) domains, which recognise and activate the monomer substrate; condensation (C) domains, which catalyse amide bond formation; and thiolation (T) domains, which shuttle reaction intermediates between catalytic domains.1 This arrangement offers prospects for rational peptide modification via substitution of substrate-specifying domains within an assembly line.2,3 For over 20 years, it has been considered that both A domains and C domains play key roles in selecting and “proof-reading” the substrate;4 a presumption that has greatly complicated rational NRPS redesign.2,5 Here we present evidence derived both from natural and directed evolution studies that C domains do not play a substrate-specifying role, and that novel non-ribosomal peptides can be generated by substitution of A domains alone. We identify permissive A domain recombination boundaries and show that these allow us to efficiently generate novel pyoverdine peptides at high yields. We further demonstrate the transferability of our approach in the PheATE-ProCAT model system originally used to infer C domain substrate specificity, generating novel dipeptide products at yields that should not be attainable according to existing dogma.