real cognitive automation by deep text …2ed6568a-3198-4bb4-8228...real cognitive automation by...
TRANSCRIPT
Real Cognitive Automation by Deep Text UnderstandingNico Lavarini
Autonomous Decision-Making: Assessing the Technology and its Impact on Industry and SocietyRüschlikon, October 25, 2017
▪ LARGEST EUROPEAN VENDOR of Text Analytics & Cognitive Computing
▪ PUBLIC COMPANY (EXSY) with offices and R&D labs in Europe and USA
▪ PATENTED TECHNOLOGY
▪ The technology of choice
FOR ENTERPRISES in all sectors
and FOR GOVERNMENTS
About us
3
COGITO leverages Natural Language Understanding by means of Semantic Technology, to process documents and drive high ROI
Cognitive capabilities on artificial intelligence algorithms that mimic the human ability
10+ years of large projects worldwide Domain- and Application-independent semantic platform Full set of unstructured information (80% of total) management features
Categorization, Text Analytics, Tagging and Ontology Matching Extensive Knowledge Representation, Search Relationships, Sentiment analysis
What we do
4
To what purpose
Knowledge Management
• Automatic Classification
• Content Enrichment
• Advanced Analytics
Cognitive Automation
• Customer Support & Conversation
• Knowledge-based Business Processes
Corporate Intelligence
• Competitive Intelligence
• Operational Risk Mitigation
For what customersGOVERNMENT
FEDERAL AGENCIES
MEDIA & PUBLISHING
LIFE SCIENCE & PHARMA
BANKING & INSURANCE
TELCO & OTHERS
OIL & GAS
MANUFACTURING
6
What we do differently
From Information
to KnowledgeFull Semantic Understanding
Language Knowledge
Graphs
Cognitive Computing Platform
The COGITO Technology
7
8
Background
RPA• Clerical Process Automation
by SW robots• Rule-based, non-disruptive
Cognitive Computing• Simulating human abilities• Computer Vision, Speech
Recognition, Predictive Analysis, IoT• Text Mining, NLP, Machine Learning,
Sentiment/Emotion, Reasoning
COGITO
9
From Text Analytics to Cognitive Automation with RPA Forrester on RPA: “design the system to potentially link with cognitive
platforms.”
Started from simple cognitive processes (email routing, customer care, document categorization)
Moving to Process Restructuring with RPA with full Cognitive Computing deep impacts
Cognitive Technology and RPA
10
Process Class for Process Automation
Basic RPA
• transactional work activities• rule-based / repetitive in nature• (screen-scraping, macros, workflows
and basic design capabilities)
Enhanced• recognition of unstructured data• adapting to the business
environment
Cognitive
• decision support• advanced decision algorithms• interlinked with artificial
intelligence, NLP, analytics
Improves cost/speedData Information
+ Reasoning +ValidationInformation Knowledge
+Decision AutomationKnowledge Understanding
11
RPA vs Cognitive Automation
Task Type
High VolumeLow Complexity
RoutineStructured Data
High ComplexityNon-Routine
Decision-supportingUnstructured Data
Operational Mode
Instruction-based Needs data and human training
Disruption in Job Definition
Low to medium High
Investment
Lower CostShorter Time
High ROI
Higher CostLonger TimeHighest ROI
Cogn
itive
Aut
omat
ion
RPA
12
Case: Claims Management
•Motor, P&C in general – medical info, claim data, etc.
•High payouts, frauds, skill and turnover of Claims Managers
Case
•Health: automatic claim validation Accept/Reject (Decision Automation)
• Injury: from 1h to 1min for Medical Report analysis Price Range
Results
•Time saving, cost saving (FTEs and leakage), cohesion and coherence
•Decision automationImpacts
13
Case: Property Risk Evaluation
•Risk Engineers often evaluate and rate Risk Factors from 3rd party Risk Reports
•Risk Selection & Pricing, non standard ratingsCase
•Cost effectiveness, volume management and focus on direct Risk EngineeringResults
•Actual Risk to Price compliance•Reduction of SubjectivityImpacts
14
Case: Contract Matching
•Claim/Contract, Master/Local, Instruction/Policy
•Match, misalignments, leakages, high volumes, language barriers, local conventions
Case
• Speed up Claim Adjudication, reduce Misalignment/Leakage among contracts or UW InstructionsResults
•Reduce reserving, restructure contract processes, evaluate consistencyImpacts
15
Case: Intelligence
• Smart Alerting for Intelligence Analysts• Daily job is information search and filtering (from
IP theft to Brand Management to Security)Case
• From continuous Pull to Push and alerting –Improved work quality
• Big Data sustainable impactResults
• Time Saving, consistency, 24/7 availability• Applications to Investment Management, Business
and Competitive Intelligence• Global awareness, hidden Signals, links
Impacts
16
Time Saving & Cost Effectiveness – 24/7 availability Insurance: increase capacity/reduce reserving Reduce Risk/Leakage/Inconsistency
Coherence and cohesion – remove error-prone processes Human focus on more complex and bracing tasks added value
Note: Global quality evaluation is complex (cognitive biases) Inter-Rater Agreement (sometimes 60%) Intra-Rater Agreement (consistency over time)
General impacts
17
Highlights
• Tradeoff of expertise transfer to machines• Human acceptance (steals my job, discredits my
expertise)• Action responsibility for machine-driven decisions
On Automation
• Support to decisions and not necessarily unassisted automation
• Decision | Validation | Alerting
On Decisions
18
Outlook
“Zero-FTE Back Office”? (McKinsey: 30% can be automated) Better customer/broker experience direct business benefit Biggest impact is Business Process transformation: Getting prepared
Setting expectations Empowering Employees Education/training/dissemination/sponsoring/Focus groups
Knowledge Transfer is Key
linkedin.com/company/expert-system
twitter.com/Expert_System
Nico LavariniChief Scientist
[email protected] (+39) 329 8607 804