Post on 30-May-2022
COMPANY PROFILE
CHALLENGES
John Hancock, a large insurance provider, uses ANTstein™ to automate data extraction for claims
processing, eliminate manual tasks andincrease business productivity.
TECHNOLOGY/OPERATIONAL CHALLENGES
Documents had to be read manually and keyed into their downstream systems for further processing
Extracted data primarily handwritten (cursive) and signature, had to be verified for the accuracyof information capture
Corporate headquarters:Boston, Massachusetts, United States
Global employees: 8,000 (2018)
Revenue:US $531 Million
SOLUTIONThe AntWorks Solution
The documents were first cleaned up with the pre-processing engine then classified using the DocID engine (document indexing)
Cursive handwriting was extracted dynamically from the documents and reinforced using advanced, deep learning techniques
Exceptions were handled using the innovative quality check screen, which apart fromcorrection of data, enabled the system to learn with both assistive and adaptive machine learning
Finally, this clean data was fed directly into the downstream systems using API integrations
The large volume of policy management documents received needed to be manually handled
Policy documents held vast amounts of unstructured data, specifically handwritten text in bold and cursive
Signatures needed to be identified and validated with other documents to ensure authenticity
AntWorks was not only able to process unstructured data that included handwritten fields (both block & cursive), but also check-boxes and signatures within the policy documents and contextualise the extracted fields.
KEY OUTCOMES
ANTstein, AntWorks’ flagship integrated automation platform (IAP), has a unique ability to read cursive writing and offers:
The AntWorks Advantage
Noise reduction, using an image enhancement engine that is unaffected by watermarks, blurs or overlaps A foundation in fractal science that requires a smaller training data set
Learning based on pattern recognition and not on an absolute font set or pixel matching
FOR HANDWRITTEN CURSIVE RECOGNITION(accuracy increases with more samples and training)
65%+ACCURACY
75%REDUCTION
IN MANUAL WORKFORCEfor the tedious process of data extraction
INCREASED ACCURACYin extracting data from process documents
EASIER workflow management
LOWER turn-around time
HIGHER business productivity
High Level – Process Flow
Pre-processing
Fetch document from Client repository and enhance the
Image quality for only the “Poor Quality” documents
Document Identification
Classifies documents into respective document categories.
A “Quality Check” screen is provided to validate the
classifications
Solution Approach To Meet Client Objectives
BENEFITS
Delivery Design forNon-Linear Profitability
Optimise Time and Benefit Trade Off
Create a Scalable Content Capture Framework
Data Capture
From unstructured documents containing Printed / Handwritten /
Checkbox / Signature content
Data Enrichment
Validates and applies necessary business rules to captured data
Quality Audit
Allows user to validate the output data and complete the
transaction
Output
Final data, post-validation exported in any required format
(xls, csv, xml & json) or APIprovided for target systems to
consume the data
• Maximise Data Capture Accuracy using technology intervention
• Automate Classification using Learning Algorithms and Fractal Networks
• Utilise Machine Learning and AI to achieve data certainty in Extraction
• Analyse operational data to identify 80:20 benefit rule
• Design Implementation plan to maximise and optimise booking of
benefits while managing risks
• Shape long-term partnership for mutual benefit
THANK YOU
To find out how you can leverage ANTstein solutions for your organisation, contact us at
Hello@Ant.Workswww.Ant.Works