Why medical coding is broken and how AI can fix it

Medical coding quietly underpins reimbursement, reporting, and population health, yet the systems behind it remain painfully manual, inconsistent, and hard to scale. In this episode of The Builders Series, Corti’s Joakim and Andreas unpack why medical coding is still so difficult, from outdated classification systems and fragmented national standards to noisy training data and poor generalization across hospitals.

They explain why traditional approaches, including keyword matching, rule-based tools, and narrowly trained classifiers, keep falling short, and why the next generation of coding systems needs to “code like humans.” That means using language models to reason through guidelines, not just memorize historical patterns.

This is a conversation about infrastructure, standards, and what it really takes to build healthcare AI that works in the real world.