Corti has released Symphony for Medical Coding, a new agentic AI model designed to convert unstructured clinical notes into accurate ICD-10 codes. Unlike traditional pattern matching systems, Symphony uses a multi agent architecture that mimics the reasoning process of human coders. It is now available via API and claims to outperform major models from OpenAI, Anthropic, Google, Amazon, and Oracle by more than 25% in clinical accuracy benchmarks.
Glimpse:
The model employs a four stage agentic workflow identifying clinical evidence, interpreting coding hierarchies, validating against guidelines, and resolving ambiguities. It addresses challenges from longer clinical documentation driven by ambient AI scribes. Symphony has been validated on benchmarks like ACI-BENCH and MDACE, as well as real world U.S. health system data in emergency and outpatient settings. It also demonstrated the ability to identify three times more suicide attempts than traditional coding in a Danish patient data study.
Corti, a clinical grade AI laboratory founded in 2016, has launched Symphony for Medical Coding, an advanced agentic AI model aimed at transforming medical coding workflows. Medical coding involves translating complex, unstructured clinical notes into standardized ICD-10 codes a system with approximately 70,000 diagnosis codes used for billing, research, and healthcare policy.
Traditional coding automation tools often rely on pattern recognition from annotated datasets, which can struggle with rare codes, multiple specialties, or frequent guideline updates. Symphony takes a different approach by encoding coding rules directly into its architecture and using a multi agent system to reason over clinical text in a way that closely replicates how trained human coders work.
The model follows a four stage process: agents identify relevant clinical evidence, navigate coding hierarchies, validate entries against official guidelines, and resolve any ambiguities. This reasoning-based method improves accuracy and explainability, especially as ambient AI scribes generate longer and more detailed clinical notes.
In independent benchmarks (ACI-BENCH and MDACE), Symphony outperformed general purpose models from leading tech companies including Claude Opus 4.6, GPT-5.4, and Gemini 3.1 Pro by more than 25% in clinical accuracy. The results were consistent across multiple test runs. Real world validation on data from a large U.S. health system in emergency and outpatient settings further confirmed its performance.
In one study using Danish patient data, the model identified three times as many suicide attempts as were previously recorded through conventional coding, highlighting its potential to uncover important clinical insights that might otherwise be missed.
Symphony for Medical Coding is now available through Corti’s API and integrates with the company’s Agentic Framework. It supports HIPAA and GDPR compliance, with growing demand from U.S. customers for strong data privacy safeguards. The solution targets electronic health record vendors, virtual care platforms, practice management systems, and life sciences organisations.
“Symphony for Medical Coding is built differently. It encodes the rules of coding directly into the architecture and uses a multi agent workflow to reason over those rules against the clinical text.”
By
HB Team
