The last five years have seen rapid adoption of automation and artificial intelligence (AI) in the medical coding space. Hospitals, physician practices, and revenue-cycle vendors all promise faster cycle times, lower labor overhead, and improved accuracy — but the reality is nuanced. Below I examine what autonomous (fully automated) and human-assisted (AI + human reviewer) coding mean in practice, summarize documented error-rate data, and lay out the practical benefits, risks, and implementation best practices organizations should consider. Definitions: what we mean by “autonomous” and “human-assisted” coding Autonomous coding describes systems that ingest clinical documentation (structured EHR data and/or unstructured clinical notes) and automatically produce ICD-10, CPT/HCPCS, modifiers, and claim bundles …