AI-Powered Healthcare Triage & Claims Automation
Challenge
The healthcare provider network experienced long emergency department triage wait times, increasing left-without-being-seen (LWBS) rates, and frequent insurance claim denials. Clinicians had to review fragmented patient records across multiple Electronic Health Record (EHR) modules, creating documentation inconsistencies and increasing administrative workloads.
The organization needed an AI-powered solution to support nurse-led triage, summarize clinical notes, and improve claims coding efficiency while preserving clinician oversight. The implementation also needed to integrate with existing healthcare systems without disrupting daily operations.
Strict security and compliance requirements added further complexity. Patient health information needed appropriate protection, including de-identification, controlled processing, auditability, and safeguards against exposing sensitive data. The goal was to improve patient flow and revenue cycle performance while maintaining human review of AI-generated clinical and coding recommendations.
The healthcare provider network experienced long emergency department triage wait times, increasing left-without-being-seen (LWBS) rates, and frequent insurance claim denials. Clinicians had to review fragmented patient records across multiple Electronic Health Record (EHR) modules, creating documentation inconsistencies and increasing administrative workloads.
The organization needed an AI-powered solution to support nurse-led triage, summarize clinical notes, and improve claims coding efficiency while preserving clinician oversight. The implementation also needed to integrate with existing healthcare systems without disrupting daily operations.
Strict security and compliance requirements added further complexity. Patient health information needed appropriate protection, including de-identification, controlled processing, auditability, and safeguards against exposing sensitive data. The goal was to improve patient flow and revenue cycle performance while maintaining human review of AI-generated clinical and coding recommendations.
Solution
Viston developed an integrated healthcare AI solution combining an AI triage assistant, clinical summarization, and claims coding automation. The project began with discovery, workflow assessment, and privacy evaluation to define integration requirements and establish safeguards for Protected Health Information (PHI).
The nurse-facing triage assistant supported symptom intake and acuity scoring to help standardize assessments. An LLM-powered summarization system converted fragmented clinical notes into structured SOAP summaries and generated ICD-10 suggestions for clinician review.
For revenue cycle operations, a claims coding assistant identified missing documentation and supported CPT/ICD code recommendations through a human-in-the-loop review process. The solution integrated with existing EHR workflows through Epic APIs and FHIR interfaces.
The technology stack included Azure OpenAI Service, Presidio for PHI de-identification, FastAPI, Azure Kubernetes Service, Azure Cognitive Search, Redis, PostgreSQL, MLflow, Evidently AI, Power BI, and Terraform. Privacy safeguards, audit trails, model monitoring, and controlled human review supported responsible deployment across the healthcare network.
Viston developed an integrated healthcare AI solution combining an AI triage assistant, clinical summarization, and claims coding automation. The project began with discovery, workflow assessment, and privacy evaluation to define integration requirements and establish safeguards for Protected Health Information (PHI).
The nurse-facing triage assistant supported symptom intake and acuity scoring to help standardize assessments. An LLM-powered summarization system converted fragmented clinical notes into structured SOAP summaries and generated ICD-10 suggestions for clinician review.
For revenue cycle operations, a claims coding assistant identified missing documentation and supported CPT/ICD code recommendations through a human-in-the-loop review process. The solution integrated with existing EHR workflows through Epic APIs and FHIR interfaces.
The technology stack included Azure OpenAI Service, Presidio for PHI de-identification, FastAPI, Azure Kubernetes Service, Azure Cognitive Search, Redis, PostgreSQL, MLflow, Evidently AI, Power BI, and Terraform. Privacy safeguards, audit trails, model monitoring, and controlled human review supported responsible deployment across the healthcare network.
Results
The healthcare AI solution improved triage efficiency, claims processing, and patient digital experience across a seven-hospital network.
Average triage time decreased from 41 minutes to 29 minutes, representing a 28% reduction. The left-without-being-seen (LWBS) rate fell from 5.2% to 3.7%, indicating improved patient flow. The claims workflow also improved, with the reported claim denial rate decreasing from 9.4% to 8.3% and claim processing turnaround time dropping from 9.2 days to 6.0 days.
Overall, claim denials decreased by 12%, claim turnaround improved by 35%, and patient digital experience increased by 18 Net Promoter Score (NPS) points.
The implementation included a $580,000 investment covering the minimum viable product and production hardening. The MVP was delivered in 16 weeks, followed by a full rollout completed over 28 weeks across seven hospitals.
These results demonstrate how AI-assisted triage, clinical summarization, and claims coding can help healthcare organizations reduce administrative burden and improve operational performance while retaining human oversight and privacy safeguards.
The healthcare AI solution improved triage efficiency, claims processing, and patient digital experience across a seven-hospital network.
Average triage time decreased from 41 minutes to 29 minutes, representing a 28% reduction. The left-without-being-seen (LWBS) rate fell from 5.2% to 3.7%, indicating improved patient flow. The claims workflow also improved, with the reported claim denial rate decreasing from 9.4% to 8.3% and claim processing turnaround time dropping from 9.2 days to 6.0 days.
Overall, claim denials decreased by 12%, claim turnaround improved by 35%, and patient digital experience increased by 18 Net Promoter Score (NPS) points.
The implementation included a $580,000 investment covering the minimum viable product and production hardening. The MVP was delivered in 16 weeks, followed by a full rollout completed over 28 weeks across seven hospitals.
These results demonstrate how AI-assisted triage, clinical summarization, and claims coding can help healthcare organizations reduce administrative burden and improve operational performance while retaining human oversight and privacy safeguards.