Why medical coding remains a challenge in modern practices

Dealing with systemic coding challenges

The challenges of medical coding are not new. What is new is the level of complexity practices are being asked to manage amid rising financial strain and persistent resourcing questions.

Increasingly, coding accuracy is less about selecting a code and more about translating clinical documentation into the correct level of specificity. As documentation becomes more detailed with the rise of ambient scribes, voice dictation, and other AI-powered note capture methods, the volume of information has increased. But the risk of inconsistency, ambiguity, and misinterpretation have increased alongside that volume.

For experienced teams, the issue is not a lack of knowledge or attention to detail. It’s that coding sits at the intersection of several moving targets: clinical documentation quality, code set complexity, regulatory nuance, and constant time pressure. Each of these elements evolves independently, and rarely in ways that simplify the work.

As a result, coding has become increasingly more dependent on clinical context. At the same time, coding itself has become more complex. There are tens of thousands of codes, each with specific documentation requirements, nuanced distinctions, and frequent updates that require ongoing training and auditing.

In some cases, incomplete visibility into prior documentation or supporting context can add friction. However, more often the challenge lies in how documentation is captured and interpreted — not just whether it is available.

That is why many coding challenges today are systemic rather than individual. Errors and denials are often the result of misalignment across systems and processes, not isolated mistakes. Even well-trained, well-supported teams encounter friction because the environment itself is inherently complex.

What’s changed in medical coding — and why traditional fixes are no longer enough

Historically, practices have addressed coding challenges by adding more oversight, such as more training sessions, more audits, and more staff dedicated to reviews. Those approaches still play a role, but they are no longer keeping pace with the scale of the problem.

One of the biggest shifts is the sheer volume of documentation. Ambient scribes and AI-assisted tools are helping to enable more detailed capture of clinical encounters. From a coding perspective, however, it can also introduce new ambiguity that must be interpreted and validated. As a result, coding and billing teams spend more time validating work that has already been completed.^1-4

This shift shows up most clearly in rework. Appeals, corrections, and resubmissions to payers are consuming a growing share of high-value staff time, while introducing delays and variability into reimbursement. Over time, that pressure can contribute to burnout of not only revenue cycle staff but also providers.

On a day-to-day basis, this often shows up as longer work queues, growing backlogs, delayed claims, and more handoffs between clinical, coding, and billing teams. Work that could move forward instead gets paused and revisited, which adds friction across the entire revenue cycle.

There’s also something to be said about the sheer number of medical codes. The American Medical Association announced in Sept. 2025 the addition of 288 new codes for the Current Procedural Technology code set. On top of that, there were 84 code deletions and 46 code revisions. The regular updating and sheer nuance across tens of thousands of codes necessitates regular auditing because of the amount of risk involved.

The “why now” is not a single new regulation or requirement. It is about the cumulative effect of increasing documentation volume, changes to code sets, and operational strain. Incremental improvements alone are no longer enough to manage the coding complexity for a growing provider office.

Where AI fits in medical coding and where it doesn’t

As practices look for new approaches, artificial intelligence (AI) is often positioned as a broad solution. In reality, its value in medical coding is more targeted and more practical. AI is not replacing coders or clinicians. Instead, it is helping shift where their time and judgment are applied.

AI can be highly effective in high-volume, repeatable scenarios where patterns are consistent, and rules are well-established. It can help identify inconsistencies between documentation and selected codes, flag missing elements, and highlight potential compliance or coding risks before a claim is submitted.

In these cases, the value is not automation for its own sake, but earlier visibility — or outright prevention — of issues that would otherwise require rework. Practice leaders might consider how well their current medical coding and billing processes work in high-volume environments, and whether they have software with built-in rules engines that can help enable AI.

That evaluation might include AI’s ability to help maintain both speed and quality by consistently applying established coding logic across large volumes of documentation — something that is difficult to sustain through manual review alone. Not to mention, AI also has the capacity for auditing code sets quickly and flagging any codes for revision or deletion.

At the same time, there are clear boundaries. When documentation is ambiguous, or when clinical context requires interpretation, human expertise remains essential. These are the situations where judgment, not pattern recognition, determines the right outcome. Understanding this distinction is critical:

• AI is strongest in identifying patterns, omissions, and risks across large volumes of data.
• AI is weakest when context is incomplete or requires nuanced interpretation.

When applied in the right places, AI does not replace decision-making, but instead helps to focus it. Knowing that distinction can help practice leaders evaluate where to introduce AI first within specific workflows.

How AI-supported coding changes the day-to-day experience for teams

The impact of AI-supported coding is often most visible not in metrics, but in how work feels day to day.

For coding and billing teams, the shift is subtle but meaningful. Instead of spending time correcting issues after submission, the teams can address potential problems earlier in the workflow. That helps reduce the volume of rework, limits the need to revisit completed claims, and creates a more predictable rhythm of work. They can also leverage AI for consistent auditing purposes to keep up to address any regulatory or compliance concerns.

For clinicians, the experience changes in a different way. Documentation still matters, but the burden shifts. Instead of trying to anticipate how documentation will be interpreted later, clinicians receive clearer signals about whether their documentation supports accurate coding as AI helps issue flags and provide additional details to help improve coding rather than introducing new inconsistencies. 

Sending those signals early is imperative as documentation volume increases because it offers structured support that can help improve documentation, which can in turn help improve coding. That helps change where cognitive load sits by reducing the mental overhead associated with second-guessing or anticipating rework and offering a proactive response instead of reactive.

Over time, this can lead to smoother clinical workflows, with fewer interruptions from coding queries and less need to revisit completed encounters, allowing clinicians to stay more focused on patient care rather than downstream administrative implications.

When systems provide clearer signals about what supports accurate coding, clinicians can focus more on capturing relevant clinical detail and less on anticipating how someone will later interpret the medical codes.

The alignment between documentation, coding, and billing is often where practices see the greatest operational relief. Better-connected workflows can help reduce friction across both clinical and revenue cycle tasks.

The financial implications leaders should pay attention to

While AI is often discussed in terms of cost savings, the more meaningful impact for practice leaders is how it changes financial stability. Predictability can be just as important as total revenue. When claims are cleaner and denials are reduced, reimbursement becomes more consistent. That consistency helps support better planning, more reliable cash flow, and less dependence on reactive fixes.

In many cases, preventing a denial can have a greater impact than accelerating a submission. It avoids the downstream costs of appeals, reduces delays, and limits the operational strain placed on staff.

These dynamics can also influence workforce stability. When teams spend less time on repetitive corrections and rework, the risk of burnout decreases, and practices are better positioned to retain experienced staff.

What to evaluate before introducing AI into coding workflows

Adopting AI in medical coding is not simply a matter of adding new technology. It requires a clear understanding of how work currently flows through the organization. Before introducing AI-supported tools, practices should take a step back and evaluate:

• Where denials are originating, such as documentation gaps, inconsistencies in coding training and quality assurance, or downstream billing issues
• Which coding tasks are repetitive and pattern-driven, versus those that require judgment
• How feedback is communicated back to clinicians
• How they will use the AI outputs

This kind of assessment helps practices apply AI where it can add value, rather than introducing new complexity. It is crucial to evaluate readiness and align new capabilities with existing workflows.

How athenahealth supports AI-assisted coding

The athenahealth approach to AI-assisted coding reflects a broader design philosophy: support decision-making without removing control.

Rather than introducing separate tools or parallel processes, AI capabilities are embedded within existing workflows. This allows clinicians and staff to benefit from additional insight without having to change how they work.

The emphasis is on:

• Maximizing both speed and accuracy, rather than sacrificing one for the other
• Delivering context, rather than opaque outputs
• Integrating into familiar workflows, rather than creating new ones

This approach aligns with the athenaOne® AI-native evolution, where intelligence is built directly into the EHR and revenue cycle processes instead of being layered on top.

Where Coding Advice and Express Coding fit into that approach

Within this framework, forthcoming athenaOne medical coding capabilities help support different points in the coding workflow.

• athenaOne Clinical Documentation Improvement (CDI) will provide in-app nudges to help clinicians close documentation gaps and ensure a seamless transition to coding and claim creation.
• athenaOne Express Coding uses deterministic and rules-based automation, supported by generative AI, to auto-populate codes based on the clinical documentation. The solution reduces manual effort while preserving quality and transparency for the user. Closer alignment between documentation and coding can help improve both accuracy and efficiency.
• athenaOne Coding Advice surfaces timely insights to support rework — and ultimately resubmission and revenue recovery — of claims denied for incorrect coding. Users trust and accept AI-generated advice 40% more frequently than human-generated advice and have seen a 26% increase in payment recovery for coding-related denials.⁵
• Together, these capabilities are not meant to replace human judgment. They are designed to reduce friction and support more consistent outcomes across the clinical documentation and medical coding process.

A more sustainable path forward for medical coding

Medical coding will always require expertise. The complexity of care, the frequency of coding updates, and the importance of accurate documentation demand deep working knowledge of the ever-changing healthcare ecosystem.

What is changing is where expertise is most valuable. AI makes it possible to shift attention away from repetitive, high-volume tasks and toward the areas that require interpretation and judgment. It helps practices manage growing complexity without simply adding more effort or overhead.

The goal is not to eliminate all the challenges of coding, but to make them more manageable, and to create systems that better support the people responsible for getting it right while allowing them to reallocate their time where their expertise can have even more value.

Practices that succeed in this environment will not rely on technology alone. They will pair it with thoughtful workflows, clear oversight, and a willingness to adapt as both care delivery and reimbursement continue to evolve.