AI Insurance & Claims Processing with Chinese LLMs

# Life insurance underwriting risk assessment import requests applicant_data = { "age": 38, "gender": "male", "occupation": "Software engineer, sedentary work", "annual_income_yuan": 420000, "coverage_requested_yuan": 2000000, "medical_history": ["Hypertension (controlled, on medication 2 years)", "No surgeries", "Family history: father had heart attack at 65"], "lifestyle": {"smoking": False, "exercise": "2x/week gym", "alcohol": "occasional", "bmi": 26.5}, "driving_record": "Clean, no accidents in 10 years", "existing_policies": "None" } response = requests.post( "https://tokenease.io/v1/chat/completions", headers={"Authorization": "Bearer YOUR_TOKENEASE_API_KEY"}, json={ "model": "deepseek-v4", "messages": [ {"role": "system", "content": "You are an insurance underwriting AI. Assess risk profiles, calculate risk scores (1-100), recommend premium adjustments, and suggest medical requirements. Comply with insurance regulations. Format as structured JSON."}, {"role": "user", "content": f"Assess underwriting risk: {json.dumps(applicant_data, ensure_ascii=False)}"} ], "max_tokens": 2000 } ) print(response.json()["choices"][0]["message"]["content"])

# Auto insurance claim damage assessment from photos import requests, base64 with open("car_damage.jpg", "rb") as f: img_b64 = base64.b64encode(f.read()).decode() claim_data = { "claim_id": "CLM-2026-88421", "policy_type": "Comprehensive auto", "vehicle": {"make": "BYD", "model": "Han EV", "year": 2024}, "incident_description": "Minor collision at intersection, front bumper and left headlight damage", "claimant_statement": "Other driver ran red light, hit my front left side", "police_report": "Filed, other driver cited" } response = requests.post( "https://tokenease.io/v1/chat/completions", headers={"Authorization": "Bearer YOUR_TOKENEASE_API_KEY"}, json={ "model": "deepseek-v4", "messages": [ {"role": "system", "content": "You are an auto claims assessment AI. Analyze damage photos, estimate repair costs, identify parts needing replacement, and flag inconsistencies with the claimant statement."}, {"role": "user", "content": f"Assess claim: {json.dumps(claim_data)}"}, {"role": "user", "content": f"Damage photo: data:image/jpeg;base64,{img_b64}"} ], "max_tokens": 2000 } ) print(response.json()["choices"][0]["message"]["content"])

# Health insurance claim fraud risk assessment import requests health_claim = { "claim_id": "HC-2026-44521", "claimant": {"age": 34, "policy_tenure_months": 8, "previous_claims_12m": 3}, "diagnosis": "Chronic lower back pain requiring physiotherapy", "treatment_cost_yuan": 28000, "provider": "Private clinic, established 2024", "claim_pattern": { "frequency": "4th claim in 8 months", "treatment_types": ["physiotherapy", "acupuncture", "massage therapy"], "provider_history": "All claims from same clinic", "referral_source": "Self-referred, no GP referral" }, "red_flags": [ "Policy purchased 2 weeks before first claim", "Clinic opened same month as policy purchase", "No prior medical history of back issues" ] } response = requests.post( "https://tokenease.io/v1/chat/completions", headers={"Authorization": "Bearer YOUR_TOKENEASE_API_KEY"}, json={ "model": "glm-4", "messages": [ {"role": "system", "content": "You are an insurance fraud detection AI. Assess claim risk, identify suspicious patterns, calculate fraud probability, and recommend investigation actions. Format as structured JSON."}, {"role": "user", "content": f"Assess fraud risk: {json.dumps(health_claim, ensure_ascii=False)}"} ], "max_tokens": 2000 } ) print(response.json()["choices"][0]["message"]["content"])

# Personalized insurance policy recommendation import requests customer_profile = { "age": 29, "family_status": "Married, expecting first child", "occupation": "Marketing manager", "annual_income_yuan": 180000, "monthly_budget_yuan": 1500, "existing_coverage": ["Basic employer health insurance"], "concerns": ["Hospital costs for childbirth", "Child education savings", "Income protection if unable to work"], "risk_tolerance": "Moderate" } available_policies = [ {"name": "Comprehensive Health A", "monthly_premium": 680, "hospital_coverage": "100% private room", "maternity": "Full coverage after 12-month waiting"}, {"name": "Term Life 20Y", "monthly_premium": 420, "coverage_yuan": 1000000, "riders": ["Critical illness", "Accident"]}, {"name": "Child Education Savings", "monthly_premium": 800, "maturity": "18 years", "guaranteed_return": "3.5%"} ] response = requests.post( "https://tokenease.io/v1/chat/completions", headers={"Authorization": "Bearer YOUR_TOKENEASE_API_KEY"}, json={ "model": "qwen3-235b", "messages": [ {"role": "system", "content": "You are an insurance advisor AI. Compare policies, explain tradeoffs, and recommend optimal coverage combinations within budget. Be transparent about exclusions and waiting periods."}, {"role": "user", "content": f"Recommend policies for: Customer={json.dumps(customer_profile)} | Policies={json.dumps(available_policies)}"} ], "max_tokens": 2000 } ) print(response.json()["choices"][0]["message"]["content"])

# Property claim narrative generation import requests property_claim = { "claim_id": "PC-2026-7721", "policyholder": "Zhang Residence", "incident_date": "2026-08-20", "incident_type": "Water damage from burst pipe", "damaged_items": [ {"item": "Wooden flooring (living room, 35 sqm)", "age_years": 5, "estimated_value_yuan": 28000}, {"item": "Built-in kitchen cabinets", "age_years": 3, "estimated_value_yuan": 15000}, {"item": "Sofa (leather)", "age_years": 2, "estimated_value_yuan": 8000} ], "cause": "Pipe burst due to freezing temperatures, building management delayed response 6 hours", "photos_submitted": True, "police_report": False, "reserve_recommended_yuan": 45000 } response = requests.post( "https://tokenease.io/v1/chat/completions", headers={"Authorization": "Bearer YOUR_TOKENEASE_API_KEY"}, json={ "model": "deepseek-v4", "messages": [ {"role": "system", "content": "You are a claims documentation AI. Generate professional claim summaries, reserve recommendations, and correspondence from structured data. Ensure regulatory compliance and consistency."}, {"role": "user", "content": f"Generate claim narrative: {json.dumps(property_claim, ensure_ascii=False)}"} ], "max_tokens": 2000 } ) print(response.json()["choices"][0]["message"]["content"])

# Catastrophe risk portfolio analysis import requests portfolio_data = { "region": "Guangdong Province", "total_policies": 45000, "total_insured_value_yuan_bn": 18.5, "peril": "Typhoon (Category 3+)", "geographic_concentration": { "shenzhen": {"policies": 18000, "value_bn": 8.2}, "guangzhou": {"policies": 15000, "value_bn": 6.1}, "zhuhai": {"policies": 12000, "value_bn": 4.2} }, "building_types": {"high_rise": "60%", "mid_rise": "25%", "low_rise": "15%"}, "historical_losses": "Typhoon Hato (2017): 320M yuan in claims from this region", "current_reinsurance": "Catastrophe excess-of-loss treaty, attachment point 200M yuan" } response = requests.post( "https://tokenease.io/v1/chat/completions", headers={"Authorization": "Bearer YOUR_TOKENEASE_API_KEY"}, json={ "model": "glm-4", "messages": [ {"role": "system", "content": "You are a catastrophe risk modeling AI. Analyze insurance portfolios, model loss scenarios, calculate probable maximum losses, and recommend reinsurance and risk mitigation strategies."}, {"role": "user", "content": f"Analyze portfolio risk: {json.dumps(portfolio_data, ensure_ascii=False)}"} ], "max_tokens": 2500 } ) print(response.json()["choices"][0]["message"]["content"])