AI at Veteran Affairs
Coa at the VA
Coa has been on the ground at the VA for 3 years working across OCTO. We've also been at the forefront of AI building teams before the surge over the past year.
We've also brought together leading partners across Healthcare, benefits, and dev tooling to identify a few high leverage use cases with AI technology at the VA.
Proposed Process: Rapid AI Deployments
Hyper accelerated 3-4 months product cycles focused on high leveraged applications of AI use cases to drive impacts across specific demographics, operational processes, and customer problems.
Demographic and Use Case Discovery
Identify key demographics, most common diagnosis for Veterans, areas of efficiency and drag, or high volume data or processing bottlenecks
Data inventory and discovery
Key databases and related dated sets, specific integrations into future of healthcare AI data (Oracle and Cerner)
Scope specific 3-4 month MVP
Identify impact against KPI's important to organizations
Test, report, validate
Key reporting structure and early data correction of models, compliance testing, and regulatory checks
Use Cases
Identifying high-leverage opportunities for AI implementation across VA operations.
Healthcare and Translational AI
AI is uniquely leveraged to help transform complex scientific ideas and breakthroughs into practical and implementable medical practices that improve patient outcomes.
Current State Challenge
In current translational medicine, the output has been algorithmic, which is hard to assess impact and distill into implementable practices that move patient outcomes.
Opportunity: Better patient outcomes through faster diagnostic integration
Translational AI, a term coined by Jason Hipp who leads Coa's Healthcare AI, can be leveraged to improve time to treat and time to diagnosis for key Veteran use cases.
The goal is to accelerate the incorporation of implementable solutions and hypothesis testing to improve policy and patient diagnostic outcomes. For example, diagnostic precision in radiology, identifying new emerging key markers of diseases in high-risk demographics, and then providing a common language of measurement to healthcare and executive leaders to measure cost and efficacy.
In this way, translational AI streamlines the incorporation of new findings into procedures and processes for healthcare workers. The simultaneous reduction of the current organizational effort to identify and incorporate them allows everyone in the intersection of care to better focus on patient outcomes.
Key Performance Indicators
Time to treat
Reduce patient wait times for treatment
Time to Diagnosis
Accelerate diagnostic processes
Key Team Member
Jason Hipp
Jason Hipp leads Coa's Healthcare AI strategy and approach. He recently coined the term Translational AI and formerly held key positions at industry leading technology and healthcare companies including: lead Pathologist at Google, the Chief Digital Innovation Officer for Mayo Collaborative Services (the commercial diagnostics arm of Mayo Clinic), and the Medical Director for Biopharma Diagnostics. He was also the first pathologist hired by Alphabet and was the lead pathologist at Google, where he was the founder and Chair of the Division of Computational Pathology & Artificial Intelligence.
Veteran Benefits
The end to end Veteran benefit journey can have a dozen or more touch points from transition all the way to the end of the appeal process. AI is already being explored to help simplify and better this process for Veterans.
Overview
The VA's Office of the Chief Technology Officer (OCTO) is already modernizing benefits systems with cloud-first platforms, agile delivery, and human-centered design to reduce processing times and improve the Veteran experience (digital.va.gov, digital.va.gov). Emerging AI capabilities can now be layered onto this foundation to automate repetitive work, unlock siloed data, and provide real-time insights across the full benefits life-cycle (department.va.gov).
Opportunity: Coa Solutions in 4 months
Coa Solutions can field a cadre of Public-Trust–cleared VA technologists—engineers, product designers, and AI specialists—who can begin work immediately and stay dedicated to a single high-priority OCTO initiative for a focused four-month engagement (digital.va.gov). Working in two-week iterations and leveraging the VA's cloud-native engineering practices (digital.va.gov), the team delivers incremental releases that demonstrate value sprint after sprint while fully meeting VA security, accessibility, and compliance standards. By combining lightweight document-AI proofs with agile integration, we surface rapid wins, trim manual backlog, and create a clear path for broader AI adoption—all within that four-month window.
Key Performance Indicators
Processing Time
Reduce time from application to decision
Application Accuracy
Improve accuracy of benefit applications and reduce rejections
Appeal Resolution
Accelerate appeals process and improve outcomes
Veteran Satisfaction
Enhance overall veteran experience and satisfaction scores
Silent Failures
Proactive identification and remediation of system failures that silently impact veteran benefits processing, ensuring no veteran is left behind due to undetected technical issues.
Overview
Charles Worthington, CTO VA, identified Va.gov errors affecting 120,000+ veteran benefits claims at the House Veterans' Affairs Subcommittee on Technology Modernization on December 4, 2023. These were primarily due to unprocessed disability claims and dependencies status update failures. The VA continued to highlight the importance of its digital transformation, specifically in the proactive identification of 'silent errors' that can harm veterans and erode trust in government solutions.
The importance of this detection was highlighted including the implementation of Watchtower monitoring systems, modernization efforts, and increasing communication to Veterans.
Opportunity: AI-Driven Anomaly Detection & Intelligent Monitoring
Recent advancements in artificial intelligence offer the VA new avenues to enhance the reliability and responsiveness of its digital systems. By incorporating AI into existing workflows, the agency can unlock deeper visibility into operational patterns, support faster detection of irregularities, and strengthen the overall delivery of services. Below are several ways AI can be applied to help monitor, analyze, and proactively address issues within the VA's benefits processing ecosystem.
Outlier Identification in Claims Flows
Train machine learning models (e.g., isolation forests or autoencoders) on historical claims metadata—submission date, type, location, claimant status—to flag outliers, such as:
- Claims that are older than expected with no downstream movement.
- Unusual concentrations of failed submissions in a region or time window.
- Duplicate or incomplete entries with no resolution path.
Embedding-Based Similarity Monitoring for Submission Consistency
Use NLP models to compare incoming free-text submissions or dependency updates to known "healthy" patterns. Detect semantic drift or unexpected omissions that suggest a backend malfunction (e.g., missing data in structured fields after front-end submission).
Real-Time Health Scoring with Ensemble AI Models
Combine event logs, system health metrics, and user interaction data to produce an AI-driven "submission integrity score" per form type and region. Automatically escalate when thresholds are breached.
AI Log Triage Assistants
Use large language models to ingest logs from across VA.gov microservices and summarize root causes, speeding up diagnosis and shortening mean time to resolution.
Key Performance Indicators
Silent Error Detection
% of form submission failures detected by AI vs. manually
Timeliness
Median time to detect & alert on silent failure
Veteran Outcomes
% reduction in benefit delivery delays due to processing issues
System Reliability
% decrease in undetected data pipeline failures (monthly)
Ready to Partner with the VA?
Let's discuss how we can help the VA implement these AI solutions and drive meaningful outcomes for veterans.