Skip to main content
All Case Studies
Insurance Technology ClaimSwift (InsurTech, UK) United Kingdom 5 months

AI Claims Processing Platform for UK InsurTech

ClaimSwift, a UK-based MGA (Managing General Agent) specialising in SME property insurance, was processing 800 claims per month manually.

3.2 days (from 14 days)
Claims cycle time
61% of claims
Straight-through processing rate
4.2x improvement
Fraud detection rate
£142,000
Annual claims leakage reduction
3.1x (same headcount)
Adjuster capacity
4.6/5 (from 3.1/5)
Customer CSAT for claims

The Challenge

ClaimSwift, a UK-based MGA (Managing General Agent) specialising in SME property insurance, was processing 800 claims per month manually. Average claims cycle time: 14 days. Adjuster headcount: 12. Fraud detection: manual spot-check by senior adjusters. Claims leakage (overpayment) estimated at £180,000 per year.

Our Solution

Canny built an AI claims management platform with automated first notice of loss (FNOL), document extraction, straight-through processing for low-complexity claims, ML-based fraud scoring, and a streamlined adjuster workbench for complex claims. Integration with Lloyd's Xchanging for bordereau reporting.

Technical Architecture

Python FastAPI backend on AWS ECS

PostgreSQL for claims data, S3 for documents

GPT-4o for FNOL extraction from email/PDF and adjuster AI assistant

Scikit-learn + XGBoost fraud scoring model trained on 3-year claims history

Textract for document OCR (photos, PDFs)

React claims workbench with embedded AI assistant

Technologies Used

PythonFastAPIReactPostgreSQLAWS ECSGPT-4oTextractXGBoostScikit-learnAWS S3

Project Timeline

1

Data Analysis and Model Design

3 weeks
  • 3-year claims dataset analysis (fraud patterns, STP eligibility criteria)
  • Claims journey mapping with senior adjusters
  • Fraud scoring feature engineering
  • STP eligibility rule definition with claims team
  • API design for Lloyd's Xchanging integration
2

FNOL and Document Processing

8 weeks
  • Email and web FNOL intake with GPT-4o extraction
  • AWS Textract OCR for claim documents, photos, and invoices
  • Claims triage: STP eligible vs adjuster review routing
  • Policyholder self-service portal for document upload and status tracking
  • Integration with policy administration system (PAS) for coverage verification
3

AI Fraud Scoring and Adjuster Workbench

8 weeks
  • XGBoost fraud scoring model trained on historical claims
  • Network analysis for related-party fraud detection
  • React adjuster workbench with AI-generated claim summary
  • Reserve recommendation engine
  • Payment disbursement workflow with bank verification
4

Lloyd's Integration and Compliance

3 weeks
  • Lloyd's Xchanging bordereau generation and submission
  • FCA data handling compliance review
  • GDPR data retention and right-to-erasure implementation
  • Penetration testing by third-party security firm
  • Team training and go-live

“Our adjusters were drowning in paperwork. Now 61% of straightforward claims go through without adjuster touch — and the AI flags the fraudulent ones our team used to miss. We handle 3x the volume with the same team.”

— James Hartley, Head of Claims — ClaimSwift

Long-Term Impact & ROI

ClaimSwift reduced claims operating cost by 40%, improved policyholder NPS significantly, and used the platform as a differentiator to win a Lloyd's syndicate partnership requiring digital claims capability. The fraud detection savings alone covered the development cost in 8 months.

Want Results Like This?

Every case study started with a free discovery call. Let's explore what's possible for your business.

Book Free Consultation

More Case Studies

Marketing Technology

AI-Powered Marketing Automation Platform

GrowthMatic (B2B SaaS) · USA & Canada

47hrs Hours Saved Per Week−78% Campaign Launch Time
Government Services / Consumer Tech

AI Passport Photo Processing Platform

PhotoID Pro · UK & Europe

15,000+ Photos Processed Daily97.2% Government Acceptance Rate