Amid staffing shortages, AI becomes medical coding
Jacqueline LaPointe·2026-06-22·via WhatIs
Even competitive providers like UC Davis Health now use AI to support thinning medical coding teams, creating more tech-driven workflows and a fundamental shift in the workforce.
Despite offering California-level salaries and fully remote positions, UC Davis Health still struggles to fill critical medical coding roles.
They aren't alone. There is a national medical coder shortage of up to 30%, according to numbers cited by the American Medical Association. This shortfall has created a talent drought so severe that even well-positioned organizations must fundamentally rethink their workforce strategy. Increasingly, that means turning to artificial intelligence.
Tami McMasters Gomez, UC Davis Health's executive director of mid-revenue cycle, says her team is using AI to augment the academic medical center's coding workforce, not replace it.
Autonomous coding and AI scribes for clinical documentation are making the mid-revenue cycle more efficient. These AI-enabled tools can take some -- but not all, Gomez emphasized -- of the tasks off the plates of shrinking teams and complete them faster and more accurately than ever before.
Scope, severity of the revcycle staffing crisis
The medical coding staffing crisis is an industry-wide shortfall of skilled, certified professionals driven by an aging workforce, industry burnout and increasing documentation complexity.
Younger generations also aren't meeting the demand, according to Gomez, which the Bureau of Labor Statistics estimates at about 14,200 openings each year on average through 2033 as current coders exit the workforce.
This isn't a problem providers can afford to wait out. The shortage will likely worsen with time if programs can't drum up enthusiasm from newcomers. This deficit could also worsen claim backlogs, delayed reimbursements and revenue leakage at a time when provider organizations are managing with razor-thin margins.
The issue is especially troubling for academic medical centers like UC Davis Health, which manage medically complex patients and cases requiring specialty coding expertise, Gomez added.
"We are honestly at an advantage with regard to some of the staffing shortages," Gomez explained. "But even despite the fact that we pay California wages and it's a remote position, we still struggle to find the people that have the skill set that can come in and work at a Level 1, large academic health system."
This has led the organization to supplement with contract labor and even develop its own unique training program specifically for neurodivergent people, but emerging technologies have offered another solution.
But although until now, the industry has been experimenting with autonomous coding, revenue cycle teams now see it as a necessity as they need more coders.
"We're looking at ways that we can adopt AI to augment the labor shortage," Gomez stated. "So, things we are doing today are bringing in autonomous coding to work in more entry-level positions for areas like radiology."
Autonomous coding has been very successful at navigating repetitive, high-volume coding, Gomez added. Since UC Davis Health performs a high volume of mammograms, MRIs, CT scans and the like every day, it often requires 12 to 15 full-time equivalents to code these encounters.
When healthcare organizations just can't fill a dozen vacant coder positions, AI can successfully augment some of the staffing struggles.
This wasn't necessarily the case even two years ago, though. AI-driven coding solutions weren't the most accurate out of the gate. However, Gomez touted the higher accuracy thresholds of more recent tools and the speed at which coding could take over a majority of cases.
"Three years ago, two years ago, it was taking a year or even longer in some instances to get an accuracy threshold of 60%," she explained. "And that is the confidence level at which we still needed to stop 40% after a year of spending hours curating the data, annotating and teaching the machine learning."
More recently, that took just 7-10 business days with one vendor for autonomous coding.
Balancing the cost of AI implementation, new hires
Autonomous coding solutions have helped UC Davis Health bridge staffing gaps in certain areas, but AI isn't always the answer to supporting teams stretched thin, Gomez cautioned.
"You have to look at the budget and the cost of FTEs versus the cost of AI," she explained. "And I know budget is top of mind for most organizations right now, with some of the cuts happening with Medicaid and other things happening with denials. Organizations are really struggling from a budgetary perspective."
You have to look at the budget and the cost of FTEs versus the cost of AI.
Tami McMasters Gomez, executive director of mid-revenue cycle, UC Davis Health
In some areas, the cost of implementing and maintaining AI may outweigh the cost of simply hiring more staff.
"In California, they may be equivalent, or it may cost us more to employ five coders than it would to have AI do that work," Gomez stated. "But somewhere like Arkansas or Missouri, the labor may be a lot less in terms of what costs look like, and the AI may be more expensive."
Healthcare organizations need to look closely at the numbers when seeking AI solutions to address their staffing shortages, she emphasized.
Understanding the current limitations of AI-enabled coding is also important. For example, Gomez wasn't confident that autonomous coding could properly code inpatient hospital cases. These cases are usually high-revenue, complex encounters that can range from more straightforward visits, like a typical newborn delivery or hip replacement, to heart transplants and trauma patients who require multiple surgeries.
"And missing one diagnosis or getting one diagnosis wrong could be the difference of $40,000 in revenue," she said. "So, there's a high risk there, and I haven't seen the autonomous coding in those areas be as robust or strong as they are with radiology coding or professional coding."
AI is evolving rapidly to take over more coding areas, but it won't solve all of a healthcare organization's staffing issues. After all, humans are still essential to the success of these technologies.
Who is a medical coder in the age of AI?
At UC Davis Health, autonomous coding isn't "plug-and-play and walk-away" despite vast improvements in coding accuracy, implementation timelines and overall revenue cycle efficiencies.
"We need to have a human in the loop," Gomez stressed. "And I don't think it's going to replace coders. We will still need QA and audit processes around AI."
However, the way medical coders work will change as AI becomes core to revenue cycle teams. For Gomez, that means upskilling coders to bolster their auditing expertise and teaching them how to work alongside AI.
"There is a recommendation to have some type of AI certification or credential," she explained.
Major medical coding industry groups are already jumping on this idea, like the AAPC -- formerly the American Academy of Professional Coders -- which offers an introductory course on AI concepts for billing and coding. The Professional Medical Billers Association also has the "Certified AI Medical Coder" designation, which demonstrates a coder's proficiency in AI-driven medical coding.
But there is a major problem with any AI certification or training course, Gomez stressed.
"We are just moving so quickly, so if we developed a curriculum and then people got credentialed on that, I think it would be outdated in six weeks," she said. "That's the challenge, so we're just upskilling people in the moment and working closely with vendors to do that."
And although being proficient in AI-driven billing and coding will be key down the line, Gomez envisions a coding workforce that is more adept at auditing.
"It's a natural progression of coders," she explained. "When you think about the coding industry, a lot of coding auditors who we have today started their career in coding. And they are already auditing humans and documentation; they will be auditing AI. Having that skill, that keen eye to audit and find gaps is what we're looking for."
That being said, the question still remains: Will AI replace medical coders?
"As the technology gets more and more advanced, we will have to look at whether we need to replace the human who has been in the loop," Gomez admitted. But that answer won't be coming for some time, she said, as vendors rapidly release new AI capabilities and more players come into the burgeoning market.
Increased focus on regulating health AI will also slow this shift in the coding workforce, as some lawmakers, such as those in California, actively work to prohibit the use of AI in certain clinical scenarios. This could affect the coding aspect, even though it's mostly considered an administrative function, because of how closely tied clinical documentation improvement and medical-necessity queries are to bedside care, Gomez explained.
For now, healthcare organizations navigating the medical coder shortage must strike a delicate balance: leveraging AI where it excels while preserving human expertise where it matters most. The path forward isn't about choosing between technology and talent; it's about strategically deploying both.
Jacqueline LaPointe is an Executive Editor at Xtelligent Healthcare Media, covering revenue cycle management, healthcare payers, health policy and health IT since 2016.
