The application of machine-learning-based artificial intelligence (AI) to structural biology has revolutionised accurate structure prediction and the design of new-to-nature proteins. The structure of many proteins and protein complexes can now be predicted with near-experimental accuracy in minutes, using the latest generation tools like AlphaFold2/3, Chai-1, or Boltz-1. Within the past three years, tools capable of accurately designing new-to-nature proteins, including RFDiffusion, ProteinMPNN, and BindCraft, have been developed. We have harnessed this technology to develop an end-to-end pipeline for the de novo generation of high-affinity binding proteins. These binders are small (5-15 kDa), single-domain proteins, with an affinity for their target comparable to antibodies and impressive stability and protease resistance. Our pipeline is largely protein target agnostic, enabling binder generation to diverse proteins, using an experimental structure or computational model of the target. We have used this pipeline to generate binders that block heme-uptake by pathogenic E. coli, modulate the function of a CRISPR-Cas gene editing system, and act as potent antagonists of a G-protein coupled receptor. These binders exhibit low nanomolar affinity, potently inhibit target function, and form complexes with their target that are structurally identical to the computational design. They are highly specific for their target protein, with no off-target binding detected against a panel of related proteins. Moreover, binder generation is rapid and cost-effective, allowing us to generate high-affinity binders to a target in weeks and for ~$15,000 – significantly faster and cheaper than antibody or nanobody technologies.