Ingress / Case Studies / Claims Processing at Machine Speed

Claims processing at machine speed.

A Fortune 1000 insurance carrier deployed document intelligence and GPT-4, cutting claim cycle time by 5 days and saving $4.8M annually.

SectorEnterprise, Insurance
TypeAI
Scale8,000 Employees
The Challenge

Adjusters buried in unstructured documents.

A top-50 US insurance carrier with 8,000 employees processed millions of claims annually. Claim adjusters spent 60-70% of their time extracting data from unstructured documents: police reports, medical records, repair estimates, photos.

The Situation

Average claim took 14 days to process from submission to decision. Two-thirds of that time was adjusters manually reading documents, highlighting relevant sections, typing data into the Guidewire claims system, and writing summaries. For a claim with a police report, repair estimate, medical exam, and photos, an adjuster could spend 3-4 hours just on document handling.

The carrier had 3,200 adjusters. Each hour saved per claim, multiplied across annual volume, represented millions in productivity value.

The Opportunity

Modern document intelligence (Azure Document Intelligence) combined with GPT-4 could extract structured fields from claims documents with high accuracy, classify claim type, and summarize key findings. That structured output could be delivered directly to Guidewire and the adjuster dashboard, reducing manual extraction from 3-4 hours to minutes of review and approval.

A human-in-loop escalation workflow would flag complex, disputed, or high-risk claims for full adjuster review.

Our Approach

Document Intelligence + GPT-4 + Guidewire Integration.

We built a processing pipeline using Otonmi, Ingress's AI division, to integrate modern AI into an existing enterprise claims system.

01
Document Taxonomy & Training Data
Audited 2,000 recent claims across auto, home, and commercial lines. Identified 12 document types (police report, medical record, repair estimate, etc.). Prepared training dataset for Azure Document Intelligence model fine-tuning. Baseline extraction accuracy measured.
Months 1-2
02
Extraction Model Build
Fine-tuned Azure Document Intelligence to extract 45 key fields (claimant name, date of loss, damage type, repair cost, liability notes, etc.). Validated on 200-claim test set. Achieved 99.2% accuracy on structured fields, 94% on semi-structured sections.
Months 2-4
03
GPT-4 Classification & Summarization
Layered GPT-4 on top of extracted data. Classifies claim type (catastrophe, litigated, fraud risk, routine). Generates executive summary and recommendation. Human-in-loop escalation rules: disputes, high payout, fraud flags sent to adjusters for full review. Routine claims pre-approved for fast payment.
Months 4-6
04
Guidewire Integration & Compliance
Integrated pipeline with Guidewire claims system via APIs. Extracted data and summaries auto-populate claim record. Compliance review: audit trail, HIPAA alignment, fair lending checks. Go-live with 500-claim pilot, validation, then enterprise rollout over 3 months.
Months 6-9
The Outcomes

Speed and savings at scale.

72%

Fewer Manual Hours Per Claim

Document extraction and entry reduced from 3.5 hours to 1 hour per claim. Adjusters now spend 90% of time on decision-making, not data entry.
5 days

Faster Claim Cycle

Average claim processing time: 14 days to 9 days. Faster data extraction enabled faster decision. Customer satisfaction improved measurably.
$4.8M

Annual Adjuster Time Savings

600,000 annual claims ร— 2.5 hours saved ร— fully loaded adjuster rate. Net savings after model ops and infrastructure costs.
Tech Stack

Azure OpenAI, Document Intelligence, Guidewire.

1

Azure Document Intelligence

Custom-trained model for insurance documents. Extracts structured and semi-structured data from PDFs, scans, and images with 99%+ accuracy.
2

Azure OpenAI (GPT-4)

Classification and summarization layer. Processes extracted data, flags escalations, generates plain-English summaries for adjusters.
3

Guidewire Claims

Enterprise claims management system. Integration via REST APIs pulls documents, writes extracted data, and logs audit trail.
4

Otonmi + Python

AI platform, orchestration, monitoring, and human-in-loop workflow. Built by Ingress's AI division. Ensures data governance and compliance.
Key Lessons

Enterprise AI requires trust and governance.

Extraction Accuracy = Adoption

Adjusters immediately trust 99% accuracy. Below 95%, skepticism grows. We spent 4 weeks tuning the extraction model to build that confidence baseline. The speed multiplier only mattered if accuracy was bulletproof.

Human-in-Loop Escalation is Non-Negotiable

We never tried to auto-approve claims. Fraud flags, disputes, high payouts, and novel claim types all escalate to human adjusters. That design decision eliminated concerns about AI making decisions without oversight. Speed and safety coexist.

Compliance and Audit Trail Essential

Insurance is regulated. We built detailed logging: what model processed the claim, what data was extracted, why it was escalated, what an adjuster approved or changed. Every decision is traceable for audits and fair lending reviews.

Change Management is the Bottleneck

The AI itself took 6 months. Training adjusters and rolling out the workflow took 3 more months. Organizational readiness, not technology, was the longest pole. We over-invested in communication and incremental rollout, and it paid dividends.

Start a conversation

Tell us what's worth doing.

// 30 minutes โ†’ a written brief.

Bring the problem. We'll come back with a written brief: what to build, what to defer, and where AI actually moves the number. No deck pitches.

Emailconnect@ingressits.com
GSA MAS#47QTCA26D000K
Reply< 24 hrs