Automated Quote Ingestion for a Specialty Insurance Firm

Industry: Insurance

Platform: Azure (Cloud Functions)

Automated Quote Ingestion for a Specialty Insurance Firm

Service: AI-Powered Document Extraction

March 24, 2026

The Challenge

Our client is a fast-growing company operating in a highly specialized segment of the insurance industry. Their team receives a high volume of insurance quotes from numerous carriers, each covering a distinct set of insured risks and arriving as PDFs ranging from 5 to 200+ pages.

Extracting the relevant data from those documents and entering it manually into their system was slow, repetitive, and error-prone. The process put a real strain on operations and created bottlenecks that were hard to scale around.

What We Built

We developed a cloud-hosted extraction tool that uses an LLM to read incoming PDF quotes and pull out the key fields automatically, then presents them to the agent for review before anything gets submitted.

The workflow is straightforward: the agent receives a quote by email, uploads it to the system, and the tool surfaces the extracted key-value pairs in a clean interface. The agent reviews the results, makes corrections if needed, and confirms. It's built around a human-in-the-loop model so accuracy stays high without removing the agent from the process.

On the technical side, the system uses a hierarchical configuration scheme — default settings layered with carrier- and line-of-business-specific overrides — to handle the wide variation in quote formats across carriers. Configurable parameters include the LLM prompt, model parameters, PDF extraction and pre-processing settings, all stored as version-controlled YAML files. This makes it easy to onboard new carriers or formats without touching core logic.

The PDF processing pipeline also includes an optimization step: for documents longer than 20 pages, the system filters out irrelevant pages such as legal boilerplate and blank forms before sending content to the LLM. In practice, this typically reduces the content passed to the model to around 7% of pages beyond page 20, keeping costs low,processing fast, and content well within context window limits..

The tool is deployed as an Azure Function with CI/CD connected to the client's GitHub repository, so new versions go live automatically on each commit.

The Result

Key-value pairs are extracted with 85% accuracy across production batches and key-value pairs.

Processing cost runs at approximately $0.20 per 1,000 pages

Comparable platforms charge around $30-40 per 1,000 pages for similar functionality, making this solution roughly 150x more cost-effective

In practice, the client stays within their cloud provider's free tier and incurs no direct usage cost

Delivery included full documentation and a strategic playbook covering recommendations for future phases and guide to scaling the system as their data needs grow.

Why It Works

The accuracy comes from close collaboration with the client during development. We worked directly with their team to build document and domain knowledge into the configuration layer, rather than relying on a generic extraction approach. The result is a system calibrated to the structure and terminology of their specific quote types — not document extraction in general.

Interested in Self-Service Analytics for Your Organization?

Whether you're in healthcare or another highly regulated industry, we can implement analytics platforms that balance accessibility with governance.