■ CRITICAL RISK ■ Healthcare
The filing-and-retrieval version of this job is already gone, and AI is now coming for the coding that replaced it. What survives is auditing, compliance, and arguing with insurance companies — humans supervising the paperwork machine.
“Electronic health records don't need a human filing cabinet.”
Our AI replacement risk score — how we score jobs
Nobody in this role has touched a manila folder in years. Modern medical records technicians work inside EHR systems like Epic and Cerner: assigning ICD-10 and CPT codes to diagnoses and procedures, checking charts for completeness, chasing physicians for missing signatures, processing release-of-information requests, and keeping documentation compliant enough to survive an audit. The coding piece is the economic heart — it's what turns clinical notes into billable claims — and it's where the pressure is concentrated.
Computer-assisted coding has been nibbling for a decade; large language models turned the nibble into a bite. Software now reads clinical documentation and suggests codes with steadily improving accuracy, and autonomous coding — claims coded and submitted with no human touch — is live in production for high-volume, standardized encounter types like radiology and pathology. EHR vendors and a crowd of well-funded startups are racing to expand that footprint. Release-of-information, chart abstraction for registries, and documentation-completeness checks are similarly being absorbed into workflow automation. The routine chart, in other words, increasingly codes itself.
The durable work is everything ambiguous and adversarial. Complex inpatient stays with comorbidities, sepsis, and surgical complications produce documentation that still defeats the software and carries enormous billing consequences. Auditing — including auditing the AI's own output — is growing, because payers deploy their own algorithms to deny claims and someone has to fight back. Compliance, privacy requests, and physician queries need humans with credibility. The field's own credentialing bodies now push coders toward auditor, clinical documentation improvement, and data-quality roles, which is the professional association's polite way of agreeing with our 84: the entry-level coding seat is the one the machine is sitting down in.
Automatability: our editorial assessment of current and near-term AI capability
Autonomous coding is in production now for standardized specialties, and the vendors are working up the complexity ladder toward inpatient care. Expect routine outpatient coding roles to thin sharply through the late 2020s while audit, compliance, and clinical documentation improvement positions grow. The occupation won't vanish — but its center of gravity shifts from producing codes to checking them, with fewer total seats.
The traditional pitch — short certificate, work-from-home coding job — is getting shakier, because routine outpatient coding is exactly what autonomous coding software does first. The field still has real openings, but plan a fast route through entry-level coding toward auditing, CDI, or compliance. Going in expecting to code routine charts for twenty years is planning a career around the machine's easiest target.
Not completely, but substantially. Standardized, high-volume specialties are already being coded autonomously in production systems, and coverage expands as the software improves. Complex inpatient cases, audits, payer disputes, and compliance work keep humans essential — partly because payers use their own algorithms to deny claims, and health systems need skilled people to fight algorithm with expertise.
Move up the judgment ladder. Auditing certifications, CDI training, and deep inpatient coding expertise all convert you from code producer to code validator — the role that grows as AI volume grows. Denial management is another expanding front. If your current job is mostly routine chart coding and release-of-information processing, start the credential upgrade now rather than after the software rollout memo.
No — they're among the most exposed in healthcare, precisely because the work is informational rather than physical. Nurses and technicians who touch patients enjoy protection this role lacks. Health information management survives as a profession, but as a smaller, more senior one focused on data quality, compliance, and audit. The pure processing layer is the part being automated out.