For years, AI has promised to revolutionize Hierarchical Condition Category (HCC) coding, yet the industry remains plagued by audit risk, hallucinated captures, and fragmented workflows. The critical question remains: When does AI performance stop being an assistive experiment and start being submission-grade?

This keynote from the founders of Martlet AI moves beyond the demo to present audited benchmark results from live enterprise deployments across a major health system and payer. We’ll deconstruct the performance gap between generic LLMs and healthcare-specific language models, utilizing real-world metrics on precision, recall, and unsupported diagnosis rates. Attendees will gain a deep dive into:

• The Regulatory Benchmark: Defining the specific performance thresholds required to match human-expert accuracy for direct submission.
• Evidence Traceability: How to deliver audit-ready AI by linking every code to specific clinical evidence within unstructured notes.
• Clinical Integrity: Strategies to maximize recall without increasing compliance risk or over-coding.
• The Architecture of Trust: Why domain-specific models materially outperform general-purpose AI in high-stakes, regulated environments.

This keynote provides is a data-driven blueprint for deploying AI that meets the highest standards of clinical and regulatory scrutiny.