Ease the pain of Travel Refunds with AI
summary
Generali replaces SAP scanning with a Telegram bot that accepts smartphone photos, then classifies receipt type with TF-IDF + image analytics and extracts total and date via a Keras model with word embeddings plus engineered features (category model reaches ~90% accuracy in trials). The pipeline adds handwritten recognition, wins a Milan Digital Week Hackfest spot, cuts manual handling from minutes to seconds per claim, and is moving toward production to deliver measurable cost savings and higher employee satisfaction.
AI-written summary · about
chapters (9)
- 0:00AI travel refund project
- 3:06Slow refund processes
- 8:08OCR with Tesseract
- 13:09Receipt analysis features
- 17:30Neural networks customization
- 21:50Testing AI solution
- 25:22Multilingual support
- 26:34Implementation and deployment
- 28:23Q&A
related talks
50mWhy use AI in insurance underwriting and claims? (PRT271)
08mAI-Driven Financial Reconciliation
20mDash of AI: Intelligent Automation to eliminate repetitive tasks
22mHarnessing AI for Underwriting in Insurance
27mBuilding Generative AI features
14mAI and Human Expertise for Excellence in Insurance
31mAI For Good: Fighting Health Insurance with AI
21mBoosting Data Efficiency with AI: Insights from a Case Study