The coder shortage is here.
AI coding closes the gap.
AI medical coding uses artificial intelligence to assign ICD-10 and CPT codes from clinical documentation. With the US facing a coder shortage of up to 30%, practices are adopting autonomous coding that writes accurate codes back into the EHR, easing staffing pressure while keeping a certified coder in the loop to review.
- 30%
- national coder shortage, up to
- 14,200
- coder openings per year through 2033
- 418
- code changes in the 2026 CPT update
The shortage stopped being a forecast.
In June 2026, reporting from TechTarget put the national shortfall of skilled, certified coders at up to 30%, citing American Medical Association figures. Even organizations offering remote roles and top-of-market pay, like UC Davis Health, say they cannot fill Level 1 coding seats. The pipeline does not refill on its own: the Bureau of Labor Statistics projects roughly 14,200 coder openings per year on average through 2033 as the current workforce ages out.
The work itself keeps getting harder. The 2026 CPT update alone made 418 changes against a code set of more than 11,000 active CPT codes. The vendor response arrived in the same window: in June 2026 athenahealth laid out an RCM roadmap of more than 80 AI features, including Express Coding, which it reports automates coding for more than 51% of beta charges and fully codes nearly one third of claims, with broad availability expected in July 2026.
- 80+
- AI RCM features on athenahealth's roadmap
- 51%
- of beta charges auto-coded
- 1/3
- of claims fully coded by AI
The shortage does not stay in the billing office.
Unfilled coder seats show up as claim backlogs, delayed reimbursement, and revenue leakage at a time when practice margins are already thin. Denials compound the pressure. A 2024 Premier survey found nearly 15% of medical claims to private payers were initially denied, and AHIMA puts the cost to rework or appeal a single denial at about $25 per claim for practices and $181 per claim for hospitals. Every empty seat makes that rework slower and more expensive.
When coding falls back onto clinicians, the cost lands as burnout. The AMA names documentation and coding time pressure among the top drivers of physician burnout, and one study of recent pediatric graduates found 81% of generalists and 78% of subspecialists wished they had more billing and coding training. The work has to go somewhere, and it usually goes to the person least equipped to absorb it.
- 15%
- of private payer claims denied initially
- $25
- to rework a single denial, per claim
- 81%
- of new clinicians wanted coding training
What AI coding actually does.
AI medical coding reads clinical documentation and proposes ICD-10 and CPT codes. The meaningful difference between tools is what happens next: does the system stop at a suggestion, or does it complete the action by writing accurate codes back into the EHR for a coder to review and approve?
The accuracy bar has moved fast. Coding leaders who once waited a year to reach a 60% confidence threshold now report reaching usable accuracy in 7 to 10 business days with newer autonomous tools. That shift is what turned AI coding from a pilot into a staffing strategy.
One principle holds across every credible deployment: keep a human in the loop. AI handles volume and consistency. Certified coders move up into review, audit, and exception handling, which is where their expertise matters most.
More headcount, or more leverage.
With up to a third of coder seats unfilled, hiring alone cannot close the gap. The practices pulling ahead give their existing team AI that writes accurate codes back into the EHR, so clean claims go out faster without new headcount.
| Approach | What it does | Best for | Limitation |
|---|---|---|---|
| Hire more coders | Add certified staff to keep pace | Complex inpatient and specialty cases | Up to 30% of seats cannot be filled; slow and costly |
| Computer-assisted coding | Suggests codes for a coder to enter | Practices wanting light assistance | Stops at a suggestion; the manual entry work remains |
| Autonomous AI coding with write-back | Assigns ICD-10 / CPT and writes them back into the EHR for review | Practices clearing high volume with a lean team | Needs a tool tuned for accuracy and EHR-native write-back |
Clean claims, with the team you already have.
CarePilot is a clinical action platform that works inside athenaOne. For coding, it does not stop at a draft: it assigns ICD-10 and CPT codes and writes them back into discrete EHR fields at 98% coding accuracy, ready for a coder to review and approve. Clean claims go out faster while certified coders shift to higher-value review, and practices go live in 1 to 2 days.
- 98%
- coding accuracy
- 78
- minutes back per day
- 1 to 2
- business days to go live
AI coding questions.
What is AI medical coding?
AI medical coding uses artificial intelligence to read clinical documentation and assign ICD-10 and CPT codes. The most capable tools write those codes back into the EHR for a certified coder to review, rather than stopping at a suggestion.
How bad is the medical coder shortage?
Reporting in June 2026 put the national shortage of certified coders at up to 30%, citing AMA figures, with the BLS projecting roughly 14,200 openings per year through 2033 as the workforce ages out.
Is autonomous medical coding accurate?
Accuracy has improved sharply. Coding leaders who once needed a year to reach a 60% confidence threshold now report usable accuracy in 7 to 10 business days with newer tools. CarePilot codes to ICD-10 and CPT at 98% accuracy with a coder reviewing before submission.
Will AI replace medical coders?
No. Credible deployments keep a human in the loop. AI handles volume and consistency while certified coders move into review, audit, and exception handling, which is where their judgment matters most.
Does AI medical coding work inside athenahealth?
Yes. CarePilot works inside athenaOne and writes ICD-10 and CPT codes back into discrete fields, with practices live in 1 to 2 days.
Close the gap without new seats.
Code a visit to discrete fields at 98% accuracy. Written back into your EHR. Request a 30-minute demo.