AI Mobile App Development in London, UK

Insurance / Insurtech

A London-based insurtech (FCA-authorized, motor and home insurance, 800,000 policyholders, Series B funded at £45M) wanted to build a mobile app that transformed the claims experience. Their existing claims process was painful: policyholders reported claims by phone (average 25-minute call), submitted paper forms and photos by email, waited 5-10 business days for initial assessment, and received settlement in 15-25 business days. The claims team of 120 staff spent 60% of their time on administrative tasks (data entry, document collection, initial assessment) rather than complex claims handling. Competitors (Lemonade, by i) had demonstrated that AI-powered claims could be processed in minutes, not weeks.

We built the AI claims app over 14 weeks: (1) AI damage assessment: for motor claims, policyholders photograph vehicle damage using the app. An on-device computer vision model (trained on 200,000 annotated vehicle damage images) classifies the damage type (scratch, dent, crack, structural), affected components (bumper, door, fender, hood, windshield), and severity level. The damage assessment runs entirely on-device (no images sent to cloud), and the model provides a repair cost estimate based on UK repair rates. For straightforward claims (single-component damage under £2,000), the app offers instant settlement — the policyholder receives a repair cost offer within 3 minutes of photographing the damage. (2) Smart FNOL: for all claim types (motor, home, travel), the app guides policyholders through a conversational FNOL process — an AI-powered chat interface asks relevant questions based on the claim type, extracts structured data (date of loss, location, description, third parties involved), and creates a complete claim record. The chat handles natural language — "someone backed into my car in the Tesco car park yesterday" is parsed into structured claim data. (3) Document AI: supporting documents (repair quotes, police reports, receipts) are photographed and processed by on-device OCR with AI extraction — the app reads the document, extracts relevant data (amounts, dates, vendor details), and attaches it to the claim record. (4) Claims tracking: real-time claim status updates with push notifications. AI-generated timeline showing expected next steps and estimated settlement date. (5) FCA compliance: all AI-driven decisions (damage assessment, instant settlement offers) are transparent — the app shows how the assessment was calculated and the policyholder can dispute or request human review. Fair value assessment documentation for AI pricing features. (6) UK GDPR: on-device image processing for damage assessment, DPIA completed, transparent data usage explanation in-app, and ICO-compliant privacy notices. Processing on Azure UK South for data residency.

Instant settlement: 35% of motor claims settled within 3 minutes of damage photo submission (no human involvement). These were straightforward, low-value claims that previously took 15-25 business days and consumed the most staff time relative to their value. FNOL time: from 25-minute phone call to 4-minute app interaction. Claims processing: average time from FNOL to settlement reduced from 18 business days to 5 business days for all claims (including complex ones — the AI triage prioritized and pre-processed complex claims for adjusters). Customer satisfaction: NPS from +12 to +56 — the fastest NPS improvement in the company's history. Claims team: team reduced from 120 to 75 through attrition over 12 months, with remaining staff handling complex, high-value claims. Staff reported higher job satisfaction — they handled interesting cases rather than data entry. App adoption: 62% of claims submitted via app within 6 months of launch (from 0% — previously phone only). Fraud detection: the AI damage assessment flagged 280 suspicious claims in the first year (pre-existing damage patterns, inconsistent damage descriptions), of which 65 were confirmed fraudulent — saving £820,000 in fraudulent payouts. Annual savings: £4.8M (staff efficiency + fraud detection + faster settlements reducing reserve holding costs). Implementation cost: £285,000. ROI: 16.8x in the first year.