Introducing
Your new presentation assistant.
Refine, enhance, and tailor your content, source relevant images, and edit visuals quicker than ever before.
Trending searches
29
AI ENABLED INSURANCE SYSTEM
EXECUTIVE SUMMARY
EXECUTIVE SUMMARY
Ayata offers Al solutions for several mission-critical functions at large insurers
Our approaches relate all data to a CUSTOMER, not a zip code or other artificial boundaries.
Ayata’s CUSTOMER-CENTRIC Al enables the expansion of CAPABILITIES from one function to another
Ayata also has Al solutions for a faster, more efficient TRANSFORMATION of Legacy Systems at a FRACTION of the COST
How Insurance Works
Distribution
Legacy System Modernization
Claims Management
Underwriting and Pricing
LEGACY SYSTEM MODERNIZATION
Challenges & Failures
Legacy Transformation with AI
- Lack of control and visibility
Uniform Customer Centric Database
- Antiquated technology - unable to scale
Industry Challanges
Automatic, Secure Process Transformation and Data Migration
- Stale and unreliable data leading to inefficient operations
- Security vulnarabilities
Uniform Customer Centric Database
Policy holder as the center of business interest
Common data platform for all insurance backend systems
Reduction in redundant and obsolete data; over 30% redundancy reduction in one core process i.e., premium calculation
Generated Accurate Policy Premium Summaries, using BrighterFuture
100% accurate policy premiums along with summaries generated for more than 12,000 active policies
Over 40% reduction in source code for premium calculation
Migrated Data Automatically from Source to Target System, using BrighterFuture
With 100% accuracy
Modern Systems
Legacy Systems
Legacy
technologies
AI
technologies
Generalized customer centric
data model
Complex standalone
data model
Ayata Solution
Complex
change management
Adaptive and
flexible
Scalable and easy to integrate and
communicate with other systems
Siloed and
unscalable
Fundamental and shared data model
across systems
Redundant and
obsolete data
Legacy System Modernization - Ayata's 2-Step Approach
Transformation Workflow
2
1
Build uniform database structure common for
all Insurance Backend Systems
Automate core processes
using AI technologies
Uniform Database Example
Process Transformation
Uniform Customer Centric Database
Virtual Insurance Backend System (IBS) Workflow
Basic Insurance Backend System (IBS) and Data Model
Convert each legacy/traditional IBS, transforming functions and associated data
Function conversion
Scale up
Meet data source requirements
Convert to virtual IBS and add new functionalities as required
Identify
functionalies
Initiate basic IBS structure and associated data model
Select and assess a legacy/traditional IBS
Data mapping and migration
Identify associated data elements
Migrate to virtual IBS and add new data elements as required
Add legacy/traditional IBS to the basis IBS
All legacy/traditional systems
Full fledged virtual IBS
Final virtual IBS with associated database
Automate Core Processes Transformation
3
1
Finally, automate remaining processes (if any) to transform the entire system
Start with the critical foundational building block required by many processes – such as Insurance Premium Calculation
Transform, all common core processes (process by process), so that they can be shared by every legacy backend system
2
Project Timeline
End-to-End transformation
for one core process
Uniform customer centric database
- Design and Structure
Months 7-9
Months 4-6
Months 1-3
Implementation Timeline
Uniform customer centric database
- Deployment
System and process architecture
- Deployment
CLAIMS
MANAGEMENT
CLAIMS MANAGEMENT
Claims Litigation Support
Medical Bill Straight Through Processing
Disability Insurance
Disability Insurance Benefit Calculation
Disability Insurance
Disability market in the US is $22 Billion, projected triple globally in the next decade
Complexity of disability claims has real business consequences
4/1 ratio of overpayment to underpayment
Transform math/judgment problems to an arithmetic problem using AI
Simple Claims
Straight-Through Processing
Digitally Enabled Customer Support
Ayata Claims Management – End-to-End Automated Solutions
Client's Internal Databases
Policy Data
Claims Data
(excluding workflows and applications)
UI Design Elements
Structured Data (policy related)
Unstructured Data (claim related)
- Basic information related to policy holder and policy documents
- Arranged and stored in tables in existing client databases
- Policy holder name, gender, DOB, occupation, etc.
- Applicable policy document number
- Claim number
- Incurred date (date of incident)
- Policy terms and conditions
- Underlying benefit definitions
- Underlying benefit payment calculation formula etc.
- Customer uploaded claim related documents in various format (PDF, Scanned Images, .txt, .word, etc.)
- Case manager controlled data in various format (phone call audio/transcript, case notes, etc.)
- Financial: tax return forms, payslips, profit & loss statement, etc.
- Medical: doctors certification, diagnosis, hospital admission, incident reports, etc.
- Others: employer's statement, miscellaneous documents
Data & API layer
Benifit Calculation
Payment
Benifit Assesment
Other Additoinal
Features
Claim Lodgment
(Sourcing Policy
T&C Service)
AI Information
Extraction & Retrieval,
Document Verification
- Patents cover methods, and are not limited to specific modules
- Cover any techniques, components, workflows for performing the steps listed above
Ayata Patent Coverage
User Interface
- Designed to capture every stage of the claim process, from initial claim reporting to final payment
- Interactive questionnaires for customer to fill out, the answer will be saved and used to lodge and assess the particular claim benefits
- U.I. displays the results from every stage of the claim process, giving direct visual of the entire automation process
- Edit buttons enable human intervention and correction when extraction and calculation results from the A.I. model have low confidence
User Interface
Benefit Lodgement Module
- Primarily structured data
- Sourcing applicable policy information, policy terms and conditions
- Automatically fetch relevant data from Policy Admin System (PAS) based on basic information such as policy number, incurred date, etc.
- Case manager can view and manage the progression of policy series in a consolidated page.
Benefit Lodgment Module
Document Extraction Module
extracted output to system database layer
Required input data schema / structure for
benefit assessment and calculation
Document Extraction Module
Keywords extraction
OCR
Image Processing
Other Information Extraction
- Employment letter
- Email communications
- Progress reports
- Miscellaneous documents
- Doctor's certification
- Medical diagnosis
- Doctor's notes
- Hospital admission/discharge
- Other medical documents
- Individual tax return
- Payslips
- Profit & loss statements
- Other payments
- Other financial documents
Example of Machine-Printed Financial Document Information Extraction
"Individual Tax Return 2014"
"(Summary)"
"1 July 2013 to 30 June 2014"
"Your name"
"Title": "MR"
"Surname or family name": "Thapa"
"Given name": "Suresh Kumar"
"Your sex": "Male"
"Has any part of your name changed since completing your last tax return": "N"
Financial Information Extraction
07/01/2013 - 06/30/2014
Tax file number: 401 203 951
Family name : Thapa
Given name : Suresh Kumar
Sex : Male
Address: 53/2A Brown Street,
ASHFIELD, NSW, 2131
Date of Birth: 10/31/1967
BSB number : 062105
Account number : 10550630
"Your postal address": "53/2A Brown Street, "
"Suburb or town": "ASHFIELD"
"State": "NSW"
"Postcode": "2131"
"Country": "Australia"
"Has your postal address changed since completing your last tax return": "Y"
Behind the Scenes
"Your date of birth": 31/10/1967
"Electronic funds transfer (EFT)"
"BSB number":"062105"
"Account number": "10550630"
"Account name": "Suresh K Thapa"
(with 90-95% accuracy)
(over 95% accuracy)
Financial Example
(with 90-95% accuracy)
(over 95% accuracy)
Example of Handwritten Medical Document Information Extraction
Self-trained deep learning model
Pre-trained HTR model
(attention encoder/decoder, transfer learning, etc)
Medical Information Extraction
Deep Learning Model
Behind the Scenes
Medical Examples
(with 85-90% accuracy)
Text Analysis and Information Retrieval
Sample Insurance Policy Document (PDF)
Passage
Machine-Readable Text
Sample Insurance Policy Document (PDF)
Question
What does the definition of total disability depend on?
Text Analysis (NLP)
language models
(BERT, BioBERT, RoBERTa, NER, Similarity, etc)
Answer
Benefit Assessment Module
Basic Benefit Assessment
(Total Disability / Partial Disability)
Ancillary Benefit Assessment
- Input: extracted information from financial and medical documents
- Assess against relevant / applicable benefit definitions from policy terms and documents
- Determine benefit type eligibility
- Output: True / False for each specific benefit.
- If True, proceed to calculation for this particular benefit
- If False, skip this benefit
Benefit Assessment Module
Optional Benefit Assessment
Offset Detection
Benefit Calculation Module
Pre-Disability Income
Basic Benefit Amount
Total Disability Benefit
Ancillary Benefit Amount
Final Payment
Partial Disability Income
Optional Benefit Amount
Offset Amount
Benefit Calculation Module
- Combination of types of benefit amount gives the Final Payment for this particular claim
- Calculate each of the benefits based on the components
- Calculate against business rules and formula
- Financial Analysis done based on all relevant information input (policy and claim)
- Benefit eligibility is also considered
- Individual components are calculated
Additional Features Relevant to Claim Management and Processing
- Real Time Push Notification: if a newly uploaded information is considered as new evidence to the calculation, the model will automatically rerun, and updated results will be pushed to the UI to notify user with such updates
- T+n & T-n modules: allows user to go to both ongoing and historical periods to re-assess a claim based on new evidences provided. System will automatically re-calculate.
Additional
Features
- Most favorable policies: based on different versions of policies, choose the most favorable results for the customers.
- Other relevant features to improve claim management and processing.
Distribution
Lead Optimization - Automated Solutions
DISTRIBUTION
- How to target potential customer to maximize the chances of purchasing the product?
- Utilizing all available information about the customer, Ayata's product allowed one company to triple its close ratio (% of people who respond to marketing end up purchasing).
Business Impact
Look-alikes in the US
Responders
Conversions
$40M Premium
Underwriting and Pricing
Ultimate Loss - Automated Solutions
UNDERWRITING AND PRICING
- Ultimate Loss Forecasts
- Coverages - Bodily Injury (BI), Collsion (COL), Personal Injury Protection (PIP), etc
- Pure Premium Forecasts
- State or State Group
- Ultimate Loss Development
- Monthly, Quarterly, Annually
- Customized Segments such as Demographical, Behavioral and Geographical
Underwriting and Pricing
Predicting Bodily Injury Loss Development
BI Loss Development - Actuarial Methods vs. Ayata