Why Human-Centered AI Matters in Modern ICD-10 Coding

ICD-10 coding is undergoing a period of sustained pressure driven by regulatory change, workforce shortages and increasing documentation complexity. As healthcare organizations seek to maintain accuracy and compliance while managing productivity demands, interest in AI-supported coding solutions has grown rapidly. But adoption remains cautiously paced, particularly where automation is perceived as a risk to quality or oversight. 

These tensions set the stage for discussion during a recent Netsmart webinar, Coding Reimagined: How AI Supports Accuracy, Compliance and Workforce Confidence. The session examined current challenges facing coding teams and offered a practical explanation of how augmented intelligence—not automation in isolation—can support coders without undermining regulatory compliance or professional judgment. 

 

The current state of ICD-10 coding in home care 

In 2026, a sort of perfect storm looms over coding teams across the industry.  

As I mentioned in the webinar, there’s regulatory complexity, constant updates to CMS rules, payer-specific requirements and regulations—and they’re always changing. Something is always new. Something is always different. 

This environment places significant cognitive and operational strain on coders, particularly as staffing shortages persist. Manual chart review remains time-intensive, often requiring coders to sift through lengthy records to identify relevant documentation. The cumulative effect is burnout risk, slower turnaround times and increased exposure to compliance errors. 

These challenges are not isolated to a single care setting or organization size. Rather, they represent systemic pressures that encourage new approaches rather than incremental process improvements. 

 

Clarifying the role of automation in ICD-10 coding

A central theme of the webinar was the need to clearly distinguish between automation and augmentation. We landed on a belief long held by Netsmart and McBee that AI should not be positioned as a replacement for professional coders. 

“AI doesn’t replace coders,” Amy Sheasby, Manager of OASIS Quality Assurance, McBee confirmed. “It’s not a replacement. It’s meant to assist.” 

This distinction is critical in a field where clinical context, regulatory interpretation and professional judgment remain essential. Rather than automating final decisions, AI can function as a support mechanism—analyzing documentation, identifying relevant information and presenting suggested codes for coder review. 

Designating this approach as augmented intelligence is a deliberate choice. As Sheasby explained, “When people hear artificial intelligence they often think automation or replacement. Augmented intelligence enhances. That distinction is central to our design philosophy.” 

By positioning AI as a supportive tool, organizations can work toward reducing manual effort while preserving coder accountability and oversight. 

 

The AlphaCoding approach to augmented intelligence

One key differentiator in AI-assisted coding is how a system is built to align with regulatory standards—specifically, CMS guidelines. 

Rather than relying on generalized AI models trained on broad datasets, purpose-built coding solutions incorporate CMS coding guidance directly into their logic. This is an intentional design: when guidelines are embedded into how the system works rather than treated as an external reference, it supports consistency and helps coders stay aligned with regulatory expectations, even as those requirements evolve. 

Another important consideration is whether or not an AI tool learns from previously coded charts. Many AI models do just that—but this approach carries significant risk. When a system relies on prior coding to inform future recommendations, it absorbs not just patterns but also errors, documentation gaps and outdated assumptions. Over time, those inaccuracies can compound, making them harder to detect and harder still to correct. 

A more responsible approach is to evaluate each record independently, grounding every recommendation in the documentation at hand and current guidelines alone. This mitigates the risk of propagating historical inaccuracies and facilitates coding decisions reflect what the record actually says. 

In a field where accuracy directly affects reimbursement, audit exposure and patient outcomes, these architectural decisions are the foundation of trustworthy AI. 

 

See how AlphaCoding puts these principles into practice. 

REQUEST A DEMO 
 

Supporting coders while strengthening compliance 

Throughout the session, we returned to the importance of maintaining coder confidence and control. AI-generated suggestions are intended to support decision-making, not override it. Coders review recommendations, apply their expertise and retain responsibility for final coding outcomes. 

As Sheasby summarized, “Our augmented intelligence tools reinforce regulatory accuracy rather than sidestepping it.” 

This emphasis reflects a broader recognition that sustainable adoption of AI in healthcare depends on trust—both in the technology and in the processes surrounding its use. Transparency, explainability and human oversight were repeatedly identified as prerequisites for successful implementation. 

Augmented intelligence offers a path forward that balances efficiency with accountability. By supporting coders rather than replacing them, organizations can address today’s operational challenges while preserving the accuracy and compliance that ICD-10 coding demands.