A I A N D I N S U R A N C E F R A U D
5 J U N E 2 0 1 9
2Capgemini Invent 2019. All rights reserved |
Three emerging trends are shaping the industry’s efforts in insurance fraud detection and prevention globally
Know Your Customer
Insurers are adopting stricter know your customer (KYC) practices for customer
identification and risk assessment
Top-10 Technology Trends in P&C
Insurance 2019
Top-10 Trends in Life Insurance
2018
Towards risk-prevention
The insurance industry is moving from the traditional risk-transfer model to a
risk-prevention model to prevent losses and reduce claims costs
It’s all about data
Insurance companies are increasingly leveraging customer behavioural data from multiple sources such as mobile, connected devices and social media
Source: Capgemini Invent
3Capgemini Invent 2019. All rights reserved |
A record number of fraud attempts were revealed in 2018 in Norway
498
336 354
428
314
201820152014 2016 2017
-11%
243 319 313 283 286
758 666 745 815 871
20172014 20182015 2016
1 001 9851 058 1 098
1 157+4%
General insurance Life and health insuranceFraudelent claims detected in MNOK
Source: Finans Norge and Capgemini Invent
FRAUDULENT AMOUNT DETECTED IN MNOK NUMBER OF FRAUDULENT ATTEMPTS DETECTED
4Capgemini Invent 2019. All rights reserved |
Given the assumption that every tenth claim is fraudulent, a large share of fraud attempts remain undetected
PERSONAL INSURANCE MARKET
TOP 3 TYPES OF GENERAL INSURANCE 2018
FRAUD IN INSURANCE
TOP 3 TYPES OF PERSONAL INSURANCE FRAUDS 2018
...a total of 827 attempts of fraud on personal general insurance was detected in
2018 = 112.3 MNOK**
214 detected fraud attempts
TravelMotor vehicles
300 detected fraud attemptsHome content
234 detected fraud attempts
UNRECORDED FRAUD
3.8m insurance deals
Travel
3.5m insurance deals
House and home content
3.4m insurance deals
Motor vehicles
Insurance experts discuss
that 10% of all
paid claims arefraudulent
=
Undetected fraudin personal
general insurance2018 is estimated
to around 2.2bn
NOK***
Number of claims (2018)
PERSONAL GENERAL INSURANCE
= approx. 22bn* NOK in pay-outs
Estimated pay-outs in bn NOK (2018)
Source: Finans Norge and Capgemini Invent
…consequently as much as 95% of all fraudulent claims might be undetected
The fact that insurance companies only detect the tip of the iceberg of fraudulent claims may be a consequence of the current approach of managing each claim isolated, manually and with minimal contextual data.
*This includes estimated insurance pay-outs for private motor vehicles, private homes incl. cabin and home content, travel and leisure boat**In addition to 44 attempts on non-personal general insurance and 286 attempts on health and life insurance, which adds up to a total of 313.8m NOK; *** Around 22bn NOK was paid out as a result of claims on personal general insurance in 2018 and the insurance industry discuss that around 10% of these are fraudulent= 2.2bn NOK
334 023
402 882684 198
12 219
Travel House and home content Motor vehicle Boat
2,2
8,9
10,4
0,5
5Capgemini Invent 2019. All rights reserved |
The most frequent type of fraud is fictious loss, while peoplein the age group 20 to 39 are the most frequent fraudsters
0
20
40
60
80
100
120
140
160
180
Fictious loss event
Exploitation of actual
loss
Altered sequence of events
Arranged loss
Inaccurate or altered
timing
Number of detected claims in 2019
PERSONAL GENERAL INSURANCE
MAIN TYPES OF FRAUD WHO COMMITS INSURANCE FRAUD?
68% 32%
MOST DETECTED FRAUD
ATTEMPTS ON TRAVEL AND
MOTOR VEHICLE INSURANCE
AGE GROUP 20-29
DETECTED FRAUD
ATTEMPTS SPLIT BETWEEN
MALE AND FEMALE
MOST DETECTED FRAUD
ATTEMPTS ON LIFE AND
HEALTH INSURANCE
AGE GROUP 30-39
...15% believe it is OK to attempt fraud on the insurance company, and 1 out of 3 under the age of 30 believe that it is OK to exaggerate the loss
amount…
Source: Finans Norge, Gjensidige and Capgemini Invent
NORWEGIAN PUBLIC ATTITUDE TOWARDS INSURANCE FRAUD
EXAMPLE USE CASES FOR AI IN INSURANCE FRAUD DETECTION
Insurers are using AI to monitor customer emotions during telephonic interactions to help customer service agents identify customer behavior and improve responses to customer needs.
Insurers are using AI (image analysis) to perform visual appraisal of damage while handling vehicle claims. It connects to the insurer’s back office and flags any suspicious claims for further investigation.
Insurers are using AI assistants to detect potential fraudulent claims and alert human analysts to take appropriate actions. The assistants are able to learn and update themselves with assessment of each claim.
6Capgemini Invent 2019. All rights reserved |
Why is ”AI” going wrong?C
ERTA
INTY
OF
CLA
SSIF
ICA
TIO
NC
OR
REC
TNES
S
GENERAL KNOWLEDGE
Although AI performances show an increasing ability to reach human-like precision and reliability on high complexity tasks, people outperform on general knowledge
7Capgemini Invent 2019. All rights reserved |
GENERAL KNOWLEDGE
CER
TAIN
TYO
FC
LASS
IFIC
ATI
ON
CO
RR
ECTN
ESS
AREA OF APPLICATION WITH
GUARANTEED BEHAVIOR
DOMAIN OF EXPERTISE
REQUIRED AREA OF COMPETENCE
Why is fraud detection so challenging,
even when applying more sophisticated AI technology?
8Capgemini Invent 2019. All rights reserved |
Limit area of application
- Keep your application
within limits
- Either «hard-coded»..
- .. or by self-
monitoring
Add a human into the loop
- Human veto on
decisions
- Avoid problems with
«soft knowledge»
- Bionic advice
Extend the area of expertise
- Extend competence
- Add background
knowledge
- Requires reasoning
capabilities
What can be done to prevent AI from going wrong?
9Capgemini Invent 2019. All rights reserved |
Some of the worlds most successful companies have one thing in common..
Company Name Change in Market Cap ($bn) (2009-31.12.2018)
Market Cap 2018 ($bn) (31.12.2018)
1 Amazon.com 706 737
2 Apple 652 746
3 Microsoft Corp 617 780
4 Alphabet 610 720
5 Berkshire Hathaway 368 502
6 Tencent Holdings 363 376
7 Facebook 293 374
8 JPMorgan Chase 220 320
9 Bank of America 194 238
10 Alibaba 187 355
Known knowledge graph builders
Source: Ycharts, Bloomberg and Capgemini Invent
10Capgemini Invent 2019. All rights reserved |
Effective use of AI knowledge graph technology will be a differentiating skill for insurance companies going forward
CRM POLICIES BILLING
EXAMPLE OF DATA SILOS
CLAIMS MGT.
EXTERNAL DATA (examples)
• Location• Activity
• Social media• Weather data
ENTERPRISE KNOWLEDGE GRAPH
11Capgemini Invent 2019. All rights reserved |
Knowledge graphs can connect various data in a systematic way to accumulate contextual knowledge and reasoning
COMMUNICATION
Signal processing
Emotion/Stress
detection
Natural language analysis
Language disam-
biguation
Speech generation
Knowledge graph
reasoning
Image analysis
Audio processing
WORK FLOW ENGINE / PROCESS KNOWLEDGE
ANOMALY DETECTION
12Capgemini Invent 2019. All rights reserved |
Insurance companies will benefit from a semantic network of interconnected facts for fraud analytics
Significant potential business value
Broaden contextual data collection
Develop capabilities in connecting various sources of data
13Capgemini Invent 2019. All rights reserved |
200 000employees
40+countries
120+nationalities
25 000employees
15countries
6 000employees
30global offices
MANAGEMENT CONSULTING
INNOVATION STRATEGYDIGITAL SERVICE
DESIGNINSIGHTS & DATA TECHNOLOGY TESTING
The Capgemini group is one of the worlds leading consulting companies for the private and public sector, with more than 200 000 employees
Robert HP Engels (PhD)Head of Data Science & AICapgemini Norway
Email: [email protected]: +47 995 44 481
Top-10 Trends in P&C Insurance 2019
Top-10 Trends in Life Insurance 2018
Fredrik BorchgrevinkHead of Insight-Driven EnterpriseCapgemini Invent Norway
Email: [email protected]: +47 907 31 230