From Zero to Deployed: Building the Future of Veteran Support with AI

The Mission

Veterans need answers, and counselors need clarity. The Department of Veterans Affairs (VA) handles an immense volume of inquiries—complex stories of service, benefits, and urgent needs. The old way? First-In-First-Out (FIFO). The new way? Intelligent Triage.

We set out to build CIPS (Counselor Inquiry Prioritization System): an AI-powered platform that reads, understands, and prioritizes every single inquiry the moment it arrives.

Rapid "Agentic" Development

This wasn't a months-long waterfall project. Using an Agentic AI Workflow, we went from a raw CSV of 200 unstructured case files to a fully deployed, secure, cloud-hosted prototype in under 24 hours.

We didn't just write code; we orchestrated a system:

AI Pipeline: We engineered a Python pipeline using OpenAI's GPT models to analyze sentiment, detect urgency, and predict routing categories with Explanation-Based reasoning.
Modern Frontend: We built a sleek, accessible interface using React & Tailwind, modeled after the Microsoft Fluent UI (Power Apps) standard to ensure seamless adoption for government users.
Deployment: We containerized the application and deployed it to a production-like environment with persistent storage, serving real-time requests instantly.

Concept Visualization

Neural triage in action: Visualizing the decision support flow.

The Tech Stack

We chose speed, stability, and scalability:

Backend: Python 3.11 + FastAPI (High performance, easy integration with AI libraries).
Frontend: React + Vite (Blazing fast load times).
Database: SQLite (Prototyping) → Roadmap to Azure SQL.
Design: Fluent UI (Native "Microsoft" feel for Federal compatibility).

Production Roadmap

This prototype is just the beginning. We designed CIPS with a clear path to FedRAMP High compliance:

GovCloud Integration: Moving the hosting to Azure Government (IL5).
Enterprise Data: Swapping SQLite for Azure SQL Database with Row-Level Security (RLS).
Secure AI: Transitioning from OpenAI API to Azure OpenAI Service (private instances, zero data retention).

Next Steps: Iterating Together

The code is live. The data is real. The impact is measurable.

Hands-on Demo: We are putting this directly in front of the customer today.
Feedback Loop: We built "Accept/Reject" logic right into the UI to let counselors teach the AI, improving accuracy with every click.
Scale: Ready to ingest thousands of daily inquiries.

“We aren't just building software; we're building a faster path to help for those who served.”

Built by the B&A Innovation Team.