AI coding tools promise speed and revenue. Medicomp’s David Lareau on the rework, denial and audit costs they can add.

AI-generated diagnosis and procedure codes can raise claim volume and revenue while adding rework, denials, audit exposure and unpredictable token costs that offset the gains. In an HCB News column, Medicomp Systems CEO David Lareau argues the cause is the probabilistic design of general-purpose LLMs, which can assign severity no one can trace to the record, and that a deterministic clinical intelligence layer returns the same auditable result every time. For health systems, the more useful question is whether they can prove where a code came from, not how fast a model can produce one.