---
title: "Discussion on \"How I Built an LLM-Powered Insurance Support Classifier That Saved $200K Annually\" | Hashnode"
description: "Discussion on \"How I Built an LLM-Powered Insurance Support Classifier That Saved $200K Annually\". From 3 Human Support agents manually triaging 16,000 emails/month to an $8/month AI pipeline\n\n\nAt Layr, we sell commercial insurance to small businesses. That means every day, hundreds of emails land"
url: "https://hashnode.com/posts/how-i-built-an-llm-powered-insurance-support-classifier-that-saved-200k-annually/6a245154e4962c4245787a4e"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [How I Built an LLM-Powered Insurance Support Classifier That Saved $200K Annually](https://generative-ai-journal.hashnode.dev/how-i-built-an-llm-powered-insurance-support-classifier-that-saved-200k-annually?utm_source=hashnode&utm_medium=feed)

From 3 Human Support agents manually triaging 16,000 emails/month to an $8/month AI pipeline At Layr, we sell commercial insurance to small businesses. That means every day, hundreds of emails land

Jun 6 · 8 min read · generative-ai-journal.hashnode.dev  
Read the article

## Written by

[Neelesh Roy](https://hashnode.com/@neeleshroy)

@neeleshroy

  
[View profile](https://hashnode.com/@neeleshroy)

Senior/Staff Full-Stack Engineer | LLM & agentic systems, platform & CI/CD | Python, TS/JS, React, AWS Serverless

Full-stack engineer with 10 years across frontend, backend, and cloud. I build end to end: TypeScript and React on the front, Node and Python on AWS serverless (Lambda, DynamoDB, SQS, SNS, API Gateway, Cognito) on the back. My range runs from the browser to the platform: 1. A broker-facing reporting dashboard (broker satisfaction up 24%) 2. A quote-generation frontend that automates new-business follow-ups 3. A serverless migration from a v3 framework to AWS SAM + Terraform that cut ~$20k, and the CI/CD that gates security, quality, and regression checks on every release 4. The team's Claude Code workflow and ran the training that got the whole team using it Open to senior, staff, architect full-stack roles

## Responses

Markdown supported · ⌘↵ to post

No responses yet.
