AI in Insurance Statistics: 147 Documented Deployments

147 documented AI deployments in insurance by use case, line of business, technology, insurer and vendor, with results as each source reports them.

Last updated
Maintained by
Peter KorpakLead Editor
147
Documented deployments
70
Named organizations
5
Vendors with profiles
71%
Report a quantified result

What the records show

147 AI deployments in insurance are documented here, drawn from 117 source publishers. Claims Processing is the most documented use case (49 deployments), and Predictive ML is the most common technology (49). 105 deployments (71%) report at least one quantified result; the other 42 describe an implementation without a usable figure.

Every figure below is a count of case-study records or a result quoted from its source. See the methodology and caveats before using them.

Use cases by line of business

Each deployment is counted once, under one use case and one line of business. Row and column totals add up to 147 deployments.

By use case

Reported results list up to 5 records per use case, one per named organization, chosen by rule rather than by hand: human-reviewed records first, then records that passed the automated evidence gate, higher extraction confidence, newer sources, and finally alphabetical order. Forecasts, simulations, targets, context figures, project timings and volume counts are left out. Metrics and values are not compared or averaged across studies, because each source measures something different.

Claims Processing

Deployments
49
With a quantified result
76% (37)
Named organizations
32
Top technologies
Generative AI (16), Computer Vision (15), Predictive ML (10)
Reported results for Claims Processing, as reported by the source
OrganizationMetric, as reported by the sourceValue, as reported by the sourceCase study
Admiral SegurosClaims Processed Without Appraiser Input90%Admiral Seguros achieves 90% touchless claims processing with AI-powered visual damage assessment
AllianzClaims Processing Time Reduction80%Allianz cuts food spoilage claims processing from days to hours with agentic AI
Allianz PartnersClaims Automation Rate (≤12 hours)71%Allianz Partners achieves 71% automation rate and reduces claims lifecycle from 19 days to 4 with AI
AllstateEmail Drafting Time Reduction70%Allstate cuts email drafting time 70% and halves complaint rates with GPT-powered claims communication
AXA SwitzerlandClaims Processing Cost Reduction25%AXA Switzerland Cuts Claims Processing Costs 25% with AI-Powered Straight-Through Processing Platform

Underwriting Automation

Deployments
30
With a quantified result
63% (19)
Named organizations
20
Top technologies
Generative AI (11), Predictive ML (11), NLP (7)
Reported results for Underwriting Automation, as reported by the source
OrganizationMetric, as reported by the sourceValue, as reported by the sourceCase study
American International Group (AIG)Underwriting Speed Improvement2X–5X faster end-to-endAIG Processes 370K Submissions 5X Faster with Agentic AI Orchestration
At-BayUnderwriting Decision TimeUnder 2 minutesAt-Bay accelerates cyber insurance underwriting decisions to under 2 minutes with Censys internet intelligence
AvivaUnderwriter Review Time per Case~50% reductionAviva cuts underwriter review time by 50% with AI summarisation tool for critical illness cover
AXAPolicy Review Time Reduction70%AXA accelerates underwriting analysis from weeks to hours and cuts policy review times 70% with AI automation
Corvus InsuranceLoss RatioBelow 40%Corvus Insurance achieves sub-40% loss ratio with generative AI-powered underwriting automation

Document Processing & OCR

Deployments
17
With a quantified result
71% (12)
Named organizations
10
Top technologies
NLP (9), Generative AI (4), Predictive ML (4)
Reported results for Document Processing & OCR, as reported by the source
OrganizationMetric, as reported by the sourceValue, as reported by the sourceCase study
Corebridge FinancialData Entry Time ReductionUp to 70%Corebridge Financial cuts data entry time 70% with ML-powered hyperautomation
QBE InsuranceClaim Processing Speed80% faster (documents processed in 20% of prior time)QBE Insurance cuts claim processing time by 80% with AI document automation
Shepherd InsuranceProductivity Boost270x faster than manual entryShepherd Insurance achieves 270x productivity boost in claims data extraction with OCR automation
Tokio Marine Insurance UAEClaims Processing Time Reduction90%Tokio Marine Insurance UAE cuts claims processing time 90% with AI-driven document automation
AllianzEmployee Hours Returned Monthly10,000 hoursAllianz automates 2.5 million transactions annually with intelligent RPA digital workforce

Customer Service & Chatbots

Deployments
16
With a quantified result
88% (14)
Named organizations
12
Top technologies
NLP (7), Generative AI (6), Predictive ML (3)
Reported results for Customer Service & Chatbots, as reported by the source
OrganizationMetric, as reported by the sourceValue, as reported by the sourceCase study
AllstateFirst-Call Resolution RateImproved from 67% to 75%Allstate reduces call duration 9% and improves first-call resolution to 75% with AI digital colleague
AvivaNet Promoter Score Improvement7x increaseAviva cuts liability assessment time by 23 days and boosts NPS 7x with AI-powered claims transformation
AXACustomer Renewal AHT Reduction23% (182 seconds in 2 months)AXA reduces average handle time 23% and supplier call volume 34% with Verint AI-powered analytics
Generali FranceCalls Resolved Without Human Intervention1.3 million calls (30% of requests)Generali France resolves 1.3 million calls autonomously and processes 2.1 million operations with Microsoft Copilot and agentic AI
Generali PolandClose Rate Increase24%Generali Poland increases close rates 24% with AI-powered website individualization

Fraud Detection

Deployments
13
With a quantified result
69% (9)
Named organizations
9
Top technologies
Predictive ML (7), Generative AI (5), Computer Vision (1)
Reported results for Fraud Detection, as reported by the source
OrganizationMetric, as reported by the sourceValue, as reported by the sourceCase study
AND-E (Aioi Nissay Dowa Europe)Fraud Detection Improvement120%AND-E achieves 120% improvement in fraud detection with continuously learning AI model
AXA SwitzerlandQuery & Processing Time ReductionOver 95% (from days to minutes/seconds)AXA Switzerland cuts query times 95% and generates double-digit million CHF profit with Google Cloud AI transformation
MAPFREFraud Detection Rate (Recall)+31%MAPFRE improves homeowner insurance fraud detection by 31% using synthetic data
Progressive InsuranceCampaign Lift (AI on vs. AI off)197%Progressive Insurance drives 197% campaign lift and faster ML insights with Generative AI and H2O.ai
Zurich Insurance GroupStraight-Through Processing Rate50%+ for certain personal lines segmentsZurich Insurance Group deploys 160+ AI use cases across claims, underwriting, and fraud detection

Customer Acquisition & Retention

Deployments
5
With a quantified result
60% (3)
Named organizations
4
Top technologies
Predictive ML (4), Computer Vision (1)
Reported results for Customer Acquisition & Retention, as reported by the source
OrganizationMetric, as reported by the sourceValue, as reported by the sourceCase study
Ping An InsuranceNew Business Value Growth39.8% YoY (RMB 22.3B)Ping An Insurance deploys AI across 650 scenarios to drive 39.8% NBV growth and 70% faster claims processing
Allianz DirectCost Reduction30-50%Allianz Direct achieves 15% revenue growth and 30-50% cost reduction through AI-powered digital platform transformation

Policy Management

Deployments
3
With a quantified result
100% (3)
Named organizations
3
Top technologies
Predictive ML (2), Generative AI (1)
Reported results for Policy Management, as reported by the source
OrganizationMetric, as reported by the sourceValue, as reported by the sourceCase study
Canal InsurancePolicy Onboarding Time60% reductionCanal Insurance cuts policy onboarding time 60% with HubSpot Enterprise CRM overhaul
HDFC ERGO General InsuranceProduct Launch CycleReduced from months to 4 weeksHDFC ERGO cuts product launch cycle from months to four weeks with AI-powered core system overhaul
Mitsui Sumitomo InsurancePolicy Renewal Supplemental Coverage Rate250% increaseMitsui Sumitomo Insurance increases policy renewal contract rates by 250% with AI-powered sales support platform

Risk Assessment

Deployments
2
With a quantified result
100% (2)
Named organizations
2
Top technologies
Computer Vision (1), Predictive ML (1)
Reported results for Risk Assessment, as reported by the source
OrganizationMetric, as reported by the sourceValue, as reported by the sourceCase study
Zurich InsuranceModel Deployment Time69% reduction (26 weeks to 8 weeks)Zurich Insurance cuts model deployment time 69% with Amazon SageMaker AI MLOps platform for flood prediction
CSAA Insurance GroupRoof Assessment Speed2 secondsCSAA Insurance transforms property underwriting with Cape Analytics AI

By technology

Shares are of all 147 documented deployments. 1 deployment has no technology label.

Documented deployments by technology
TechnologyDeploymentsShare of allMost common use cases
Predictive ML4933%Underwriting Automation (11), Claims Processing (10), Fraud Detection (7)
Generative AI4430%Claims Processing (16), Underwriting Automation (11), Customer Service & Chatbots (6)
NLP3222%Document Processing & OCR (9), Claims Processing (7), Customer Service & Chatbots (7)
Computer Vision2014%Claims Processing (15), Customer Acquisition & Retention (1), Fraud Detection (1)
Telematics & IoT1<1%Pricing & Actuarial Modeling (1)
Not specified1<1%Claims Processing (1)

Most-documented insurers

Insurers with at least 3 documented deployments, with the records behind each count. 31 deployments name no organization or only a generic description, and are left out of these counts.

Insurers with the most documented deployments
InsurerDeploymentsUse casesCase studies
Lemonade8Claims Processing (5), Customer Acquisition & Retention (1), Customer Service & Chatbots (1), Underwriting Automation (1)
AXA5Pricing & Actuarial Modeling (2), Underwriting Automation (2), Customer Service & Chatbots (1)
Zurich Insurance Group5Claims Processing (1), Fraud Detection (1), Regulatory Compliance & Reporting (1), Risk Assessment (1), Underwriting Automation (1)
Aviva4Underwriting Automation (2), Claims Processing (1), Customer Service & Chatbots (1)
Nationwide4Claims Processing (1), Customer Acquisition & Retention (1), Customer Service & Chatbots (1), Underwriting Automation (1)
Swiss Re4Underwriting Automation (3), Regulatory Compliance & Reporting (1)
Allstate3Customer Service & Chatbots (2), Claims Processing (1)
Ping An Insurance3Claims Processing (1), Customer Acquisition & Retention (1), Customer Service & Chatbots (1)
PZU3Claims Processing (3)
The Travelers Companies3Claims Processing (2), Customer Service & Chatbots (1)
USAA3Claims Processing (1), Customer Service & Chatbots (1), Underwriting Automation (1)

Most-cited vendors

Vendors with a profile in this directory, by the number of documented deployments that cite them. 131 deployments (89%) have no vendor profile linked.

Vendors cited by the most documented deployments
VendorDeploymentsShare of all
Hyperscience64%
Tractable53%
Shift Technology32%
Cape Analytics1<1%
Cytora1<1%

Evidence profile

Record provenance

Deployments by record provenance
ProvenanceDeployments
Automated evidence gate passed133
Source-extracted record14

Source publication year

Source publication dates are not recorded for these deployments, so they are not broken down by year.

Source publishers

117 distinct publishers. The most frequent: insurancebusinessmag.com (5), cloud.google.com (4), hyperscience.ai (4), aws.amazon.com (3), insurancetimes.co.uk (3), microsoft.com (3), prnewswire.com (3), ainvest.com (2), blueprism.com (2), cognizant.com (2).

Methodology and caveats

  • Population: published insurance case studies in this directory, each citing a source. 147 records are counted as separate deployments. 21 near-duplicate write-ups of a deployment already in the directory redirect to one canonical record and are not counted again. Other articles about the same deployment may still be counted separately.
  • Results are source-reported and have not been independently verified. Metric labels and values are quoted as the source gives them; the page does not convert, average or rank them across studies.
  • A quantified result is a metric with a label and a numeric value that reports an achieved change or rate. Forecasts, simulations, targets or plans, industry or problem-context figures, project timings and bare volume counts (such as claims processed per year) do not count. A record with no such metric counts as not quantified, even if its text describes a benefit.
  • Selection bias: companies publish successes. Vendor case studies, press releases and trade coverage dominate the sources, so these counts describe what is documented, not how widely each technology is used or how often deployments fail.
  • Organization names are counted as written after reviewed aliases. Names that differ only by a leading “The” or a trailing word such as Group, Companies or Insurance are merged; regional subsidiaries are counted separately. Generic descriptions such as “a large insurer” are never counted. Most named organizations are insurers; a few insurtechs and technology providers are the subject of a record and are counted too.
  • Use case, line of business and technology are single labels assigned by the directory. A deployment that spans several is counted under one.

How the directory selects and checks sources.

Figures computed ; most recent record added .