Case Study

Automating Quote Submission Ingestion with OCR and AI

A mid-sized E&S insurer was receiving most quote submissions as emailed attachments in mixed formats, many of them handwritten. We built an intake pipeline that parses, extracts, and validates the data automatically, cutting manual entry by 90% and taking straight-through processing from zero to 85% in the first 90 days.

90%

See more results below ↓

Client

A mid-sized Excess and Surplus Lines insurer

Lines

Commercial auto

What We Solved

Submission intake and underwriting workflow

ENGAGEMENT

Custom build

The Challenge

Being easy for brokers to work with was central to how this company won business. It leveraged a broad appetite and a willingness to write high-risk policies to hold strong market share, and brokers rewarded that flexibility with volume.

The volume arrived by email. Most quote submissions came as attachments in whatever format the broker happened to use, frequently including handwritten elements. Each one had to be read, interpreted, and keyed into the underwriting system by hand. The result was slow turnaround, inconsistent data quality, and a large share of the underwriting team’s time spent on data entry rather than on risk.

The Solution

An end-to-end automated intake pipeline. Optical Character Recognition and a custom AI model parse the emailed attachments, then extract, validate, and inject structured data into the company’s proprietary underwriting system, where submissions can be scored and prioritized for review by the underwriting team.

Technologies

Python NestJS PostgreSQL OpenAI

Results

90%

reduction in manual data entry

85%

straight-through processing rate within the first 90 days, up from zero

95%

average field-level accuracy, and 85% on handwritten content

5x

faster initial underwriting review

The change wasn’t only speed. Because submissions arrive structured and scored, underwriters see the risks worth their attention first rather than working through the queue in the order it happened to arrive. Data quality improved as a side effect of automation rather than as a separate initiative.

Where This Applies

Submission intake is the most common first project we take on for organizations placing high volumes of business through brokers. The details differ, wholesalers, MGAs, and E&S carriers each have their own version, but the pattern is nearly always the same: submissions arriving in formats nobody controls, and people absorbing the difference.

Explore: Paktolus for Wholesalers → AI + Intelligent Automation → Underwriting + Policy Operations →

View More Case Studies

Data Engineering and Business Intelligence

A Python-based ETL pipeline replacing manual carrier data processing across more than 25 sources.

99% less reporting time
READ THE CASE STUDY →

NLP Based Error Categorization

Carrier quote rejection messages classified and routed automatically instead of by hand.

99% faster error assignment
READ THE CASE STUDY →

Automated Commission Processing

Seventy carrier statement formats mapped into parsing templates, with commissions calculated and payouts generated automatically.

50% faster commission payouts
READ THE CASE STUDY →

How would results like these impact your business?

Twenty minutes, a candid conversation. If we’re not the right fit, we’ll say so.

Let’s Talk