Freedom 2.0—Claims Automation and Mitchell’s Intelligent ...

17
1 Freedom 2.0—Claims Automation and Mitchell’s Intelligent Open Platform Olivier Baudoux SVP, Product & Artificial Intelligence

Transcript of Freedom 2.0—Claims Automation and Mitchell’s Intelligent ...

1

Freedom 2.0—Claims Automation and Mitchell’s Intelligent Open Platform

Olivier BaudouxSVP, Product & Artificial Intelligence

2

Mitchell’s Approach to Serving the Industry

Strategic PillarsAuto Physical Damage Claims Processing is rapidlychanging as a result of changes in consumerbehavior and advancements in technology. To assistour customers to keep up with changes, Mitchell’sClaim Solution strategy is focused on 4 pillars.

3

4

Gartner Says AI Technologies Will Be in Almost Every New Software Product by 2020

Today, 60% of insurers are investing in AI to improve operational processesINSURANCE | Digital Transformation Remaking an Industry, Accenture

"AI offers exciting possibilities, but unfortunately, most vendors are focused on the goal of simply building and marketing an AI-based product rather than first identifying needs, potential uses and the business value to customers."

Jim Hare, research vice president at Gartner

4 out of 5 companies are actively pursuing investments in AI, with 41% having already deployed the technology.

Artificial Intelligence has become a Must-Have

McKenzie | The Time to Act is Now

5

Machine Learning | AI timelineClaims

AutomationSince an early flush of optimism in the 1950s, smaller subsets of artificial intelligence –first machine learning, then deep learning – have created even larger disruptions.

Artificial Intelligence

Machine Learning

Deep Learning

1950s 1960s 1970s 1980s 2000s 2010 2015 2018 CURRENT

Early artificial intelligence stirs excitement.

Machine learning begins to flourish.

Deep learning breakthroughs drive AI boom.

1990s

Source: @Google

6

What can machine learning do?

Machine Learning uses data to create models that:

Can adapt to a changing environmentExample: Vehicles never seen before

Classify new data, or make predictionsExample: Image Classification, optimal parts prediction

Provide concise summaries and insightsExample: Cluster estimates, discover accuracy opportunities

Source: @UCSD

7

What AI is not? The Magic Pill to Touchless Claims

Mobile First Notice of Loss

Consumer Uploads Photos

Estimate is automatically generated from submitted photos

Consumer receives estimate

Consumer selects repair facilities

Once repair is complete, the repair facility is automatically paid

Telematics Identifies Car Accident

8

Our Journey towards Claim Automation

Develop Vehicle Damage Detection

(VDD) AI model

Leverage Computer Vision & Neural

Networks to recognize damage areas through

claim photos

AI-driven Assisted Review

Review parts and labor recommendations based

on computer vision output

Mitchell Intelligent Estimating

Leverage AI solutions to partially or fully

automate the creation of estimates

AI-Driven Smart Solutions

Expand to other capabilities to solve for

other business problems.

e.g. Smart Triage

9

Vehicle Damage Detection (VDD)IN

PUTS

Car/No Car

Completed Models

Core Mitchell IP

Mitchell Intelligent Estimating

Panel Detect Damage Detect Operation Parts & Labor

10

Strategic Partnership with Google

Mitchell - Google VDD AI Model• Enhance MVDD model to maintain competitive edge• Continuous training to extend model knowledge and

classifiers leveraging Google’s expertise• Expose VDD in new global markets

Google Cloud Platform• Optimization for Google Cloud Platform• Surfaced on Google Marketplace

11

Vehicle Damage Detection (VDD)IN

PUTS

Car/No Car

Completed Models

Under-Development

Core Mitchell IP

Panel Detect

Damage Detect

Damage Type

Damage Severity Operation Parts &

Labor

Mitchell Intelligent Estimating

12

The Foundations are Rapidly StrengtheningPanels/Components Model Progression over the last 12-months

Front Bumper

Front BumperFront Upper BumperFront Lower Bumper

Right Headlamps Left Headlamps

GrilleUpper Grille

Grille EmblemFront Bumper

Front Upper BumperFront Lower Bumper

Front Reinforcement Bar

13

Assisted Review through Computer Vision

Confidence indicated with icon

Desk Reviewers will have the option to “Agree” or “Disagree” with the photo analysis

14

Assisted Review – Efficiency & AccuracyUtilize AI to review claim photos and estimate details to flag cases with potential inaccuracies

Efficiency Accuracy

Drive focus towards claims with highest potential opportunities

Ensure the right amount is paid on each claim

15

Assisted Review Objectives & Benefits

Maximize Reviewer ThroughputMetric: Show where the final approved estimates aligned with Assisted Review results

Metric: Gauge efficiency gains by Assisted Review categorizing the images and estimate lines by panel for the reviewer

Performance ManagementMetric: Results by estimating resource/shop to identify performance issues and training opportunities

Selecting the Right Estimates to ReviewMetric: Validate results where original version estimates showed Assisted Review exceptions

Identify Estimate InaccuraciesMetric: Measure repair/replace operation exceptions compared to final decision

Measure: Identify additional damage opportunities

16

Press ReleaseFor Immediate ReleaseIntroducing…

Mitchell Intelligent Estimating™

1717

Key Takeaways

Be clear on what business problem you are trying to solve

Select your AI partners wisely – Not all AI vendors are created equal

Identify your own AI approach – Buy, Build vs. Partner

The best AI vendor is unlikely to stay the best tomorrow