Medical coding is not becoming obsolete. The U.S. Bureau of Labor Statistics projects 7% job growth for medical records specialists from 2024 to 2034, a rate classified as “much faster than average,” adding roughly 13,800 positions to the field. That said, the role is changing significantly as artificial intelligence handles more routine coding tasks, and coders who don’t adapt their skills will find fewer opportunities over time.
What AI Can and Cannot Do Right Now
About 48% of healthcare organizations now apply AI to documentation and coding, making it the leading use of artificial intelligence in hospital revenue cycles. That number is expected to climb as organizations look for ways to offset staffing shortages and improve cash flow. For straightforward cases, like a simple office visit with a single diagnosis, AI tools can suggest codes quickly and accurately.
But the technology has real limitations. A study published in NEJM AI found that large language models are fundamentally poor at matching alphanumeric medical codes to their correct descriptions. The problem is structural: AI systems process text in short segments called tokens, and when that process is applied to medical codes, it scrambles the coding system’s built-in logic. Overall digit-level accuracy was below 70% across all models tested. These aren’t minor rounding errors. A wrong digit in a medical code can mean the difference between a routine follow-up and a major surgical procedure, with enormous financial consequences.
Complex cases make things harder still. When a patient has multiple chronic conditions, when a surgeon’s operative note is ambiguous, or when clinical context determines which diagnosis is primary, AI struggles with the same subjective judgment calls that make coding intellectually demanding in the first place. As one longtime health information management expert put it: “As long as there is the element of subjectivity, clinical care, and context, I think we’ll be involved.”
Why Coding Errors Cost Too Much to Leave to Machines Alone
Claim denials cost U.S. hospitals roughly $262 billion per year, creating serious cash-flow problems across the industry. Providers fail to collect 2% to 5% of net patient revenue partly because of inefficient billing processes, including coding mistakes. In a system where a single misassigned code can trigger a denied claim, the financial incentive to keep experienced humans in the loop is enormous.
There’s also an interesting psychological dynamic at play. Research from a radiology department found that people hold AI to a stricter standard than they hold other humans. The acceptable error rate for AI was 6.8%, while the acceptable error rate for a human reader was 11.3%. That gap matters because it means healthcare organizations won’t trust fully automated coding until AI performs nearly twice as well as the human coders it replaces. We’re not close to that threshold yet, particularly for the complex, high-value claims where accuracy matters most.
ICD-11 Is Making Coding More Complex, Not Less
The global transition to ICD-11, the newest version of the international classification system, is adding complexity rather than simplifying it. The system has considerably more functions, more detailed coding steps, and more linkages between categories. A qualitative study of early implementers found that users struggle to identify the right codes and understand the expanded terminology. One implementation leader was blunt: “No one is made redundant. We need the people that we have now more than ever.”
Countries adopting ICD-11 have faced shortages of technical expertise, inadequate training infrastructure, and workflow disruptions that require experienced coders to resolve. Facilities have had to redesign their data entry systems, splitting responsibilities between clerks and clinical coders in new ways. Rather than eliminating coding jobs, the transition is creating demand for coders who can navigate a more sophisticated system and train others to do the same.
How the Role Is Shifting
The coders most likely to thrive are those moving beyond simple code assignment into roles that require clinical judgment and analytical thinking. Two areas stand out.
Clinical documentation improvement (CDI) involves reviewing physician notes before and during a hospital stay to ensure the documentation supports the most accurate codes possible. This work requires deep coding knowledge combined with the ability to communicate with physicians about what’s missing or unclear in their notes. Many CDI professionals have years of coding experience that gives them a stronger grasp of documentation requirements than staff who receive only brief training in the specialty.
Risk adjustment and HCC coding is a growing sub-sector driven by Medicare Advantage and value-based care models. Health plans, large provider groups, government audit agencies, and specialized vendors all need expert coders who can accurately capture the severity of a patient’s conditions. This work directly affects how much revenue a health plan receives per patient, making accuracy both clinically and financially critical. The demand for HCC coders has increased even as basic coding tasks become more automated.
Auditing is another natural transition. Organizations need people who can review AI-generated codes, catch errors, identify patterns in denials, and ensure compliance with payer rules. Think of it as quality control for the machines.
What This Means for Your Career
If you’re considering entering medical coding or you’re already working in the field, the core skill set is evolving. The profession’s major organizations, AHIMA and AAPC, have been clear that the future belongs to coders with intellectual curiosity and decision-making ability, not those who treat coding like looking up words in a dictionary. Computer-assisted coding tools will handle the lookup part. Your value comes from understanding clinical context, catching what the software misses, and making judgment calls on ambiguous cases.
Practically, that means building expertise in areas where automation falls short: specialty coding, risk adjustment, compliance auditing, and documentation improvement. Coders with certifications in these areas, such as the CDIP or CRC credentials, position themselves for roles that AI is least likely to absorb. Entry-level, high-volume coding of simple encounters is the segment most vulnerable to automation. The ceiling, however, is rising for coders willing to grow with the technology.
The pattern playing out in medical coding mirrors what’s happened in other fields where automation arrived: the simplest tasks get absorbed, the overall profession becomes more technical, and the people who adapt end up doing more interesting, higher-paid work than before. Total employment in the field is still projected to grow through 2034, which tells you something important. Hospitals and health systems aren’t planning for a future without coders. They’re planning for a future with different ones.