Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
© 2018 Cognizant 1
© 2018 Cognizant
March 2018
AI and Analytics to Combat Digital Mortgage Fraud
A Transformative Approach
© 2018 Cognizant 2
Agenda
The Future of AI01
Human Sciences, Neural Science, and Data Sciences02
Customer Journey for Digital Mortgage Fraud03
Approaches to Protect Against Mortgage Fraud04
Neuro-ID as a Tool to Protect Against Mortgage Fraud05
Appendix06
© 2018 Cognizant 3
The Future of AI
© 2018 Cognizant 4
AI Landscape Tomorrow
By 2018, 75% of
development
teams will include
AI functionality in
one or more
application or
service.*
By 2019, 40% of
all digital
transformation
initiatives – and
100%IoT of all
effective efforts –
will be supported
by Cognitive/AI
capabilities.*
80% of
executives
believe AI
boosts
productivity.***
75% of more than
200 business
executives said
… AI will be
implemented in
their business
within 3 years.**
*Source: IDC, “IDC Sees the Dawn of the DX Economy and the Rise of the Digital-Native Enterprise.” Nov. 1, 2016, https://www.idc.com/getdoc.jsp?containerId=prUS41888916
**Source: The Economist Intelligence Unit, “Artificial Intelligence in the Real World.” 2017. http://perspectives.eiu.com/sites/default/files/Artificial_intelligence_in_the_real_world_0.pdf***Source: Su, Leo, The Motley Fool. “Ten Stats about Artificial Intelligence that will Blow You Away.” June 19, 2016. https://www.fool.com/investing/2016/06/19/10-stats-about-artificial-intelligence-that-will-b.aspx
© 2018 Cognizant 5
Human Sciences, Neural
Science, and Data Sciences
© 2018 Cognizant 6
15%35%
50%
Enterprise AI Requires a New Perspective
AI Enterprise Platform
Capabilities & Services
WORLD VIEW OF AI COGNIZANT VIEW OF AI
100%
Delivery focus
AI Enterprise
Platform
Capabilities &
Services
Teaching &
Learning
Data
Curation &
Management
Governance
Ethics, Diagnostics, & Measurements
© 2018 Cognizant 7
Data
Science
An Evolutionary View of Intelligence Sciences
Data
Science:
Designed to
provide
insights from
the historical
debris of
human activity
Human
sciences:
Observing
human activities
is designed to
unearth the
human debris of
our activities
Behavioral
Sciences: Thick
Data allows us to
develop a deep
understanding of
behavioral
insights
Neuro
Sciences:
Neuro sciences
provided the first
inside look at
how we thought
about our
activities
Curated Intelligence:
All three unifies the
systematic
identification of why
human behavior leads
to emotionally driven
outcomes
Past
+ +
Present Tomorrow
Human
Sciences
Neuro
Sciences
Data
Science
Human
Sciences
Human
Sciences
Data
Science
Neuro
Sciences
© 2018 Cognizant 8
Unifying View of Intelligence and Systems
Observational Insights
Behavioral Insights (Thick Data)
Curated
Intelligence:
Intelligent
Experiences,
Business, and
Systems
Human Sciences
Insights from
observation
(behavior)
Neuro Sciences
Insights from the brain
(thought)
Data Sciences
Insights from data
(action)
Emotional Insights
© 2018 Cognizant 9
AI Solution FrameworkInterface
(touch, voice, gesture, typing etc.)Device
(Mobile, VR, Car, home, PC etc.)
Primary Layer of AIApplication
(Both industry and cross-industry use cases)
Closed loop systems (Self-directed AI)
(Rules defined by humans)Machine
Translation
Natural Language
Processing
Personalized
Insights
Computer
Vision
Q&A Concept
Expansion
Deep
LearningRobotic
Automation
(RPA, analytics)Algorithms
Expert systems
Data sciencePeople – both leaders and
line-level coders
Platforms,
software
tools/librariesVertical utilities, APIs
Process Middleware (connects applications to networks etc.)
Underlying infrastructure (data centers, network, cloud, etc.)
Data cleansingData management, specialized ETL
AI InfrastructureSpecialized infra needed to run AI apps
Data is used to train algorithms and algorithms refines data sets for specific actions
Analyze &
learn
Act
Sense
Aggregate
Data
Develops as AI business matures
Data
© 2018 Cognizant 10
Customer Journey for
Digital Mortgage Fraud
© 2018 Cognizant 11
Customer Journey for Digital Mortgage Fraud
The occurrence of fraud during the loan lifecycle will not change if it is a digital mortgage. With digital
mortgage and data leakages, the volume and velocity will be higher
Sales
Lead
Run
Credit
Loan
Application
Order
AppraisalOrder
Title
Upload
Borrower
Docs
Prepare
Closing
Docs
Validate
Credit
File
Close
Loan
Complete
Pre-Closing
Review
Post
Closing
Review
Wire
Funds
Processing Underwriting Closing & Funding QA
Common Fraud Types
Identity Theft & Synthetic Identity Fraud
Straw Buyer
Occupancy
Mortgage Application
Sales Contract
Credit Report
Employment & Income Documentation
Asset Documentation
Appraisal
Title
Wire Fraud
Loan Estimate/Closing Disclosure
Sales
© 2018 Cognizant 12
Approaches to Protect
Against Mortgage Fraud
© 2018 Cognizant 13
Opportunity to Further Protect Against Mortgage Fraud in Today’s Digital World
Mortgage Fraud Type Leading Fraud Protection Approach ML & AI
Identity Theft & Synthetic
Identity Fraud✓
Straw Buyer ✓
Occupancy ✓
Mortgage Application ✓
Sales Contract ✓
Credit Report ✓
Employment & Income
Documentation✓
Asset Documentation ✓
Appraisal ✓
Title ✓
Wire Fraud ✓
Loan Estimate/Loan
Disclosure✓
Mortgage Fraud Protection Maturity
Traditional Competitive Leading
• Manual review of
data across
documentation
• Example:
Matching name,
address and SS
# on pay stubs,
W2s, IRS 1040s,
1003, etc.
• Use of data
tool to assess
individual
risk(s)
• Example:
Getting income
and asset data
directly from
financial
institutions
• Use of Data and
Neural Sciences to
holistically protect
against mortgage
fraud across the
mortgage lifecycle
• Example:
Generating
property fraud risk
score via LoanSafe
Fraud Manager
© 2018 Cognizant 14
Leading Capabilities to Fight Against Mortgage Fraud
Capability Description
Application Confidence Score Use data and behavioral analytics to provide confidence
level as to the accuracy of an online/mobile loan application
(i.e., LoanSafe Fraud Manager, Neuro-ID)
Cellular Network Triangulation Triangulate mobile phone location for location verification
Tax Returns/Tax Return Validations Access IRS records to vet and validate income and
employment
Income/Asset Data Direct from the
Source (Banks/ Financial Institutions)
Validate income and asset documentation without having to
upload and collect individual statements
Employment Check Incorporate tools to instantly authenticate businesses and
individuals
Email Validation Run search on email ID with social profiles to determine
confidence level as to whether the email may not be valid
and/or was created to create a fictitious persona
© 2018 Cognizant 15
Neuro-ID as a Tool to Protect
Against Mortgage Fraud
© 2018 Cognizant 16
Neuro-ID as a Tool to Protect Against Mortgage Fraud
© 2018 Cognizant 17
Neuro-ID: Insight into the 1003 Mortgage Application
MEDIUM RISK
LOW RISK
MEDIUM RISK
LOW RISK
LOW RISK
LOW RISK
LOW RISK
HIGH RISK
LOW RISK
LOW RISK
LOW RISK
MEDIUM RISK
MEDIUM RISK
(Score Range 1-99)
HIGH RISK: 1-40 MEDIUM RISK: 41-70 LOW RISK: 71-99
© 2018 Cognizant 18
Neuro-ID Use Case: Income Verification
• Proof of Concept (POC) implementation
• Online personal loan applications
• Collected data over 4 months
• 400,000 applicants monitored
• 1,100 behavioral data points per applicant
• Mouse movements, keystroke speed, etc.
Bottom 10% were about
70% more likely to fail
income verification
© 2018 Cognizant 19
Neuro-ID Use Case: Income Verification (Cont’d)
© 2018 Cognizant 20
Industry Perspective on Fraud
Occupancy Fraud
The Borrower applies for
a loan as an owner-
occupant, and they do
not intend to occupy
• Increased default rates at times as high as
50% greater than those who truthfully state
their occupancy
• Borrower may even risk a prison sentence
• The mortgage lender could call your loan due
and payable
Straw-Buyer Fraud
A Straw Buyer is a
person who makes a
home purchase on behalf
of another person, and
does not intend to live in
or control the property
Impact
• High likelihood that the loan will go into early
payment default
• Initiator of schemes often receive higher
proceeds for over-valued properties
• Property often inflated to maximize proceeds
to participants of the scheme
• Straw buyers may be criminally liable even if
they claim to not know about the fraud
Neuro-ID
• Reveals the “Digital Body Language” of the
applicant, providing a new dimension of insight
into the intent of the applicant.
• Identify and flag potentially risky applicants who
exhibit cognitive conflict around questions
pertaining to Intent to Occupy
• Visibility advances beyond the limits of historical
data and enables clearer decision making in the
application and underwriting process
• Able to identify potential fraud around questions
regarding primary residency, as well as data
familiarity throughout the application.
• Allows for fraud rules to flag potential high-risk
applicants for Straw-Buyer Fraud
Neuro-ID’s Confidence Score highlights potential risk based on anomalous behavior, for further evaluation
© 2018 Cognizant 21
Appendix
Copyright © 2018 Cognizant
Thank youFor further details, please reach out to any of the session speakers
Dr. Jerry A. SmithVP of Data Sciences and Artificial Intelligence
Cognizant Digital Business
Tel: 484-678-3518
Email: [email protected]
Justin H. WellenSenior Manager, Banking & Financial Services
Cognizant Business Consulting
Tel: 917-572-6849
Email: [email protected]
James CraddickDirector, Business Development
Neuro-ID
Tel: 509-671-7265
Email: [email protected]
www.cognizant.com www.neuro-id.com