AI/ML Can Make a Difference

INSURANCE IN THE ZETTABYTE ERA: UNDERSTANDING THE IMPACT OF IOT, BIG DATA AND ARTIFICIAL INTELLIGENCE

Dean Hamilton SVP, IoT Products / CTO Persistent Systems

IOT BIG DATA AI / ML

“For P&C insurers, it is not the rapidly changing environment that poses the biggest threat – it’s acting on future challenges with strategies linked to the past.”

Source: The Internet of Things: Opportunity for Insurers – A.T. Kearney Inc., 2014

Gordon Moore Intel Cofounder

Nielsen’s Law of Internet Bandwidth

Containerized Microservices IoT Protocol Standards

Hadoop Distributed File Systems

Stream Processing Cloud Computing

Intelligent Agents

Retailers

People Manufactures

Homes Hospitals SENSOR DATA Vehicles Communities

Energy providers Financial institutions

1 ZB = 1021 Bytes 1 ZB = 1012 GB

“In 2016 the annual run rate for global IP traffic was 1.2 ZB per year”

Source: Cisco Systems, 2017 report

IOT BIG DATA AI / ML

Master these and your business will not be disrupted!

© 2018 Accelerite. All Rights Reserved. What is Big Data? “Big data is data sets that are so voluminous and complex that traditional data processing application software are inadequate to deal with them.”

Source: Wikipedia, 2018

Big Data Sources: Social Media The 3 “Vs” of Big Data

IoT Digital Sensor Audio

Big Data VELOCITY

Digital Digital Video VARIETY Photography VOLUME 12

• More data in the last two years than entire prior human history

• 1.7 MB of new information per person on earth created every second

• 300 hours of video uploaded to YouTube per minute

TECHNOLOGIES FOR:

• Capturing • Storing • Searching/Querying • Analyzing • Visualizing • Sharing

Neuromorphic computing

Cognitive cyber Autonomous systems security

Robotic personal Machine learning assistants The appearance of

Autonomous surgical Deep learning intelligence robotics demonstrated by machine algorithms Next gen cloud Neutral networks robotics that can learn, infer, predict and

Thought controlled Pattern recognition problem solve. gaming

Real time universal Natural language translation processing

Virtual companions Chat bots

Real time emotion analytics 15

AI can effectively identify patterns AI can respond to high velocity data in high volume/variety data that much quicker than humans to predict human analysis would miss. and prevent adverse outcomes.

What should I do about it?

When will it happen again? Prescriptive Why did it happen? Predictive What happened? Diagnostic Descriptive AI/ML Differentiation

Analytics Continuum

Supervised learning Unsupervised learning

Algorithms Algorithms learn from just identify X X training 2 2 and classify dataset similarities – labels good no right or and bad wrong

results X1 X1 answers

Many common problems need both approaches (called ”semi-supervised” learning)

© 2018 Accelerite. All Rights Reserved. AI is an “umbrella term” with “cognitive computing” representing a branch of the technology focused on a specific goal – improving human-machine interaction (natural language processing, machine vision, digital assistants, etc. )

COGNITIVE COMPUTING vs. AI • Traditional AI analyzes data in order to make or recommend a A subtle but important difference distinct decision

• Cognitive computing is more concerned with helping humans interact with and obtain a clear understanding of data - so that they can make decisions

© 2018 Accelerite. All Rights Reserved. • Real-time driving behavior data are creating behavioral/usage- based products • Collision Avoidance Systems (CAS) and Advanced Driver Assistance Systems (ADAS) reduce risk AUTO INSURANCE • Semi-autonomous vehicles have the potential to reduce driving competency • Fully-autonomous vehicles will shift liability away from the driver to the manufacturers and municipalities • Challenge: Auto insurers will need to become experts in assessing software-related risks

o Connected fire/smoke detectors that alert the fire department can cut payouts by as much as $35,000 (according to Business Insider Intelligence) o Security cameras and motion sensors can HOME INSURANCE significantly reduce chances of a burglary (according to Safeguard the world) o Water leakage sensors can identify a problem before it causes major damage

• Challenge: Insurers have a vested interest in testing and promoting products that work • Challenge: Poor user experiences are turning off consumers

o Individuals who share health data can benefit from advantageous premium adjustments o Proactive individualized lifestyle suggestion can reduce risk LIFE & HEALTH INSURANCE o AI-based health monitoring can predict problems before they happen

• Tele-health systems will reduce the cost of care • Drug compliance sensors can improve healthcare outcomes • Challenge: Insurers will need to be able to generate timely insight from streaming health data

© 2018 Accelerite. All Rights Reserved. • Workplace IoT data is invaluable o Wearables can reduce the risk of workplace injury or incapacitation o Equipment condition monitoring algorithms can improve worker safety WORKER’S COMPENSATION o Environmental monitoring sensors can ensure INSURANCE regulatory compliance o Data from all of the above can be used to adjust premiums • Challenge: How to get businesses to allow you to harness this data

IOT TELEMETRY

INSURER INSIGHTS / ALERTS IoT CLOUD

BEHAVIORAL DISCOUNTS AI CUSTOMER INSURANCE

UNDERWRITING USAGE-BASED POLICIES

Mine Operator Report for Last 10 Events APIs

Field Support Service APIs

Predictive Alerting Service

Vehicles Telematics Service APIs

Devices / Sensors

Earth Moving Vehicles: Conveyors and Crushers: Workers:

• GPS, WiFi Location • Engine condition • Wearable providing • Engine condition monitor monitoring accurate location, heart • Hydraulic sensors • Conveyor load monitor rate, temperature • Tilt sensors and accelerometer

IOT TELEMETRY • Unsafe location geo-fencing with alerts INSURER • Lockout of non-credentialed INSIGHTS / ALERTS IoT CLOUD equipment operators • Equipment misuse monitoring (speed, BEHAVIORAL DISCOUNTS tilt, location, operator hours)

• Operator health monitoring with MINE AI safety alerts OPERATOR • Predictive equipment maintenance alerts

UNDERWRITING USAGE-BASED POLICIES

Contractor Portal

• Lighting • Pools • Irrigation • Vehicle Charging Stations Reserve Fund Auditor Verification Reports Smart Community Cloud Devices / Sensors

Community Resident Homeowners Portal Associations

Safety Sensors Security Sensors Environmental Sensors

• Ambient light (for • Security cameras • Storm drain / gully walkways, parking lots) • Security patrol monitor • Pool chemistry location tracking • Air quality monitor • After-hours pool/spa motion detectors • Vehicle speed monitor

• Dynamically adjust lighting for resident IOT TELEMETRY safety • Motion-based video capture INSURER throughout community • Capture and track vehicle license INSIGHTS / ALERTS IoT CLOUD plates • Alert security patrol to unsafe driver BEHAVIORAL DISCOUNTS activity (speeding) HOA • Automatically close pool if chemicals AI are at unsafe levels COMMUNITY • Alert security of after hours pool activity • Ensure security patrols provide consistent area coverage UNDERWRITING • Alert maintenance staff to blocked USAGE-BASED POLICIES storm drains

BUSINESS INTELLIGENCE OPERATIONAL INTELLIGENCE

• Helps you to analyze what has • Helps you to analyze what is happened in the past happening NOW and what may • Batch processing (daily, weekly, happen in the NEAR FUTURE monthly, etc.) is sufficient • Real-time streaming analytics is the key

ESTABLISHED BIG DATA TECHNOLOGIES

• Hadoop Distributed File System (HDFS) • Data Processing (Apache Spark) • Stream Ingestion (Apache Kafka)

DATA SCIENCE TOOLKITS What should I do about it? • Open Source Web Services (IBM Watson, Google Tensor Flow, Apple Core ML, Amazon Polly, etc.) • Python, R, Java, C++ When will it happen again? • PredictionIo, Eclipse DeepLearning4j, Apache Mahout Prescriptive • Jupyter Notebooks Why did it happen? Predictive EMERGENT TECHNOLOGIES What happened? Diagnostic • Self-Service ML Model Management • Self-Service Event Orchestration Descriptive • Self-Service Dashboard Creation AI/ML Differentiation

Analytics Continuum 39

Data Domain

Data Use Case Ingestion

Statistics Big Data Programming Data Algorithms Visualization Mining

Mathematics ML Python Algorithms Predictive Analytics Java Perfect Data Scientist! …you need a

ML Libraries Data Perfect Data Scientist Interpretation

Enterprise OI Ops. with these skills.. Data Stack Customer Empathy

https://yanirseroussi.com/2016/08/04/is-data-scientist-a-useless-job-title/ 40

Data Domain Data Ingestion Use Cases Statistics Programming Data OI Operations Visualization ML Quality Linear Algorithms Python Predictive Regression & Analytics Modeling Java Data Mining Mathematics Inferential Data Statistics Structures

Hypothesis ML Libraries Testing Big Data Business Enterprise Algorithms Context Data Stack

Customer Data Empathy Interpretation

Data Scientist OI Data Analyst/Engineer

Design Prototype Test INNOVATE

Streaming Insights Insight Logistics Data Factory PRODUCE

YOUR INSIGHTS OFFERINGS Order Marketing Billing Taking MONETIZE

YOUR INSIGHTS PLATFORM 42

Identity Mgmt. Security Data Vulnerability & Process Governance Awareness Authentication Mgmt.

• Data lake and data • Continuous vulnerability • Identity proofing • Security process modeling, warehouse operations and assessment • Continuous risk=based testing, operations and security policies • Highly responsive threat authentication monitoring • Data architecture and management • Highly responsive threat quality management management • Guaranteed trust for • Non-repudiatable customer operational insights consent

WHY? THE GOOD NEWS • Companies won’t trust an AI with unsupervised decision making • Regulators require transparency into decision making • Machine learning algorithms (like deep learning) can be highly opaque

OI ASISSTANT ALGORITHMS WILL: • Help customers reduce risk and loss • Provide input scores for dynamic pricing MORE GOOD NEWS: adjustments How will AI/ML algorithms be • Identify patterns associated with fraud used? BI ASSISTANT (COGNITIVE COMPUTING) ALGORITHMS WILL: • Identify patterns and trends could be missed • Predict potential outcomes • Advise on pricing strategies to achieve business goals

ADD PROGRAMMING SKILLS ACTUARY STRENGTH: • Mathematics • Statistical analysis TWO TRACKS • Financial theory • Understanding of domain ADD BIG-DATA PROCESSING SKILLS

INSURANCE DATA ANALYST/ENGINEER

• Traditional risks will be more easily NOW THE BAD NEWS: managed directly by the customer (Which is not really bad news) • But customers will now need operational self-insurance help • New / less manageable risks will emerge – creating new insurance opportunities

The only certainty is that the rate of disruptive innovation will continue to increase. The 3 “Vs” of Digital Transformation

Develop a mindset focused on three things: VARIETY: Experimentation is the key to success. Be bold! VELOCITY: Move fast. Fail fast. VELOCITY

VOLUME: VARIETY Be prepared to scale out. VOLUME

© 2018 Accelerite. All Rights Reserved. • Deploy a streaming data ingestion platform (for IoT and non-IoT data) • Implement a data lake and big-data Transforming the business: processing infrastructure (on private or Upgrade IT Data Infrastructure public cloud) • Deploy high performance, big-data self- service analytics tool (for OI data analysts)

© 2018 Accelerite. All Rights Reserved. • Deploy identity proofing and continuous risk- based authentication technologies Transforming the business: • Upgrade vulnerability awareness and threat Upgrade IT Security management technologies • Employ Robotic Security Process Monitoring (RSPM) technology

© 2018 Accelerite. All Rights Reserved. • Build an insurance data science organization - with augmented programming skills and tools • Build an OI data analyst organization – with Transforming the business: tools designed to leverage AI/ML models and Upgrade Workforce Skills and Tools produce timely insight to stakeholders and customers • Empower traditional teams to consume advanced analytic insight in a self-service manner

© 2018 Accelerite. All Rights Reserved. Transforming the business: • Deploy and insights monetization platform Create and Insights-as-a- • Develop a data/insight partner ecosystem Service Product Business • Develop and insights innovation process

© 2018 Accelerite. All Rights Reserved. The only thing better than being covered for a loss is avoiding the loss entirely! For example: Offer enterprise customers risk reduction insight-as-a-service • Help your customer manage operational Transforming the business: risk continuously Offer innovative new products • Connect to customer devices and to grow the business datasets and provide timely insight to prevent loss • Offer insurance rebate incentives for risk reduction • These services will offset revenue losses from price erosion in traditional areas

Preparing for the “Zettabyte Era” is a digital transformation journey – not a destination

• Choose an experienced technology partner

• Look for early wins / opportunities

• Use an iterative implementation process – “Rome wasn’t built in a day”

• Don’t just transform technology. Transform culture!