Listen to an experienced, global panel of insurance professionals present, discuss and answer your questions on the theme of “AI & Machine Learning”.
Brought to you by The Digital Insurer and sponsored by KPMG.
3. Discussion agenda
1 Presentations:
Gary Richardson: How will insurers derive value from machine learning
Adrien Cohen: AI & motor claims assessment
Juergen Rahmel: Technical considerations for AI
Alberto Chierici: Chatbots & customer service
David Robson: Enterprise view of AI
2 Questions and Answers
3 Snap Poll – share your view
4. Questions & Answers
How to participate:
If you have a question please type into the messaging area
and send to all participants
Session format:
The moderator will use a combination of his own questions
and those from the audience
10. 10
TRACTABLE MISSION : LEARN EXPERT VISUAL TASKS
WITH ARTIFICIAL INTELLIGENCE
Past 3 years have seen fundamental
breakthroughs in computer vision via
deep learning
Deep learning systems now surpass
human accuracy in certain recognition
tasks
Tech giants (Google, Facebook…) are
applying it to generic visual recognition
tasks for consumer applications
Image classification error rate
Our mission is to identify and build commercially
disruptive applications of computer vision
Our focus is insurance claims
11. 11
TEAM OF 20 BACKED BY SILICON VALLEY VC, UNIQUELY
POSITIONNED IN MOTOR WITH MITCHELL DATASET
Raised one of the largest EU
seed rounds of 2015 from West
Coast Investor
Prof. Z. Ghahramani, head of ML
@ Cambridge both an investor
and advisor
BACKERSTEAM
Partnership with Mitchell in the
US, leading insurance claims
player
Transfer dataset of 350M images
+ estimates: enables training AI
to superhuman performance
Tractable uniquely positioned
with deep learning tech & data
PARTNERSHIP IN THE US
Founding team of 3 with previous $bn exit
R&D team of 10 with 30+ years combined
research and 1000+ citations
12. 12
PRODUCT VISION : HOW TRACTABLE AI WILL CHANGE
P&C INSURANCE
Automated
Bodyshop
Adjustment
Customer
Self Service
Generate preliminary repair estimate at FNOL from photos
Applicable to auto and home
Settle low severity claims in minutes
Flag unnecessary repair procedures from photos
Collaborative workflow with the bodyshop
Contain leakage on high volume low value claims
Total Loss
Triage
Triage between repairable and total loss at FNOL from photos
Avoid unnecessary towing operations and storage fees
Manage policyholder expectations early on in process
Automate analysis of drone footage
Elastic response to claim spike during catastrophic hail events
Maintain efficient cycle times
Roof
inspection
Hail
inspection
Count dents & measure depth from photos
Elastic response to claim spike during catastrophic hail events
Maintain efficient cycle times
1
2
3
4
5
DescriptionProduct
14. 14Dr. Juergen Rahmel
Understanding Artificial Intelligence – a Tool Box
Source: Cognitive Architectures: Research Issues and Challenges by Pat Langley,
John E. Laird and Seth Rogers
Decision Making
Planning
Reasoning
Prediction
Data Intake Processing Interaction
15. 15Dr. Juergen Rahmel
Understanding Artificial Intelligence – a Tool Box
Source: Cognitive Architectures: Research Issues and Challenges by Pat Langley,
John E. Laird and Seth Rogers
Perception
Decision Making
Planning
Reasoning
Prediction
Execution
CommunicationRecognition
Data Intake Processing Interaction
16. 16Dr. Juergen Rahmel
Understanding Artificial Intelligence – a Tool Box
Source: Cognitive Architectures: Research Issues and Challenges by Pat Langley,
John E. Laird and Seth Rogers
Perception
Reflection and Learning
Decision Making
Planning
Reasoning
Prediction
Execution
CommunicationRecognition
Data Intake Processing Interaction
17. 17Dr. Juergen Rahmel
Understanding Artificial Intelligence Integration
Example: A simple ‘Chat Bot’ solution – a talkative FAQ
Reasoning ExecutionRecognition
18. 18Dr. Juergen Rahmel
Understanding Artificial Intelligence Integration
Example: A simple ‘Chat Bot’ solution – a talkative FAQ
recognize simple
keywords
search internal
rule base
reply best possible
predefined
answer
Reasoning ExecutionRecognition
Customer:
“I had a car accident,
what to do now?”
Chat Bot:
“Call Police to record the
case. Later, please submit
case number via your
Insurers Website… ”
Data / Rule Base:
…”opening hours” Mon-Fri …..
...”special offers” Offer a/b/c...
...”accident” next steps ...
...”claim” claim process ....
19. 19Dr. Juergen Rahmel
Understanding Artificial Intelligence Integration
Example: A simple ‘Chat Bot’ solution – a talkative FAQ
recognize simple
keywords
search internal
rule base
reply best possible
predefined
answer
Corporate Network
Reasoning ExecutionRecognition
Customer:
“I had a car accident,
what to do now?”
Chat Bot:
“Call Police to record the
case. Later, please submit
case number via your
Insurers Website… ”
Data / Rule Base:
…”opening hours” Mon-Fri …..
...”special offers” Offer a/b/c...
...”accident” next steps ...
...”claim” claim process ....
20. 20Dr. Juergen Rahmel
Understanding Artificial Intelligence Integration
Example: A complex ‘Chat Bot’ solution – a conversational Advisor
Corporate Network
Reasoning
Recognition
recognize
intention
clarify intention
Communication
Customer Data Product Data
Customer:
“I am thinking about increasing
my family protection”
Customer:
“I want to buy an education
insurance”
21. 21Dr. Juergen Rahmel
Understanding Artificial Intelligence Integration
Example: A complex ‘Chat Bot’ solution – a conversational Advisor
Corporate Network
Reasoning
Recognition
recognize
intention
clarify intention
Communication Communication
identify offering
customize
offering
Planning
Prediction
Reasoning
Decision Making
Customer Data Product Data
Customer:
“I am thinking about increasing
my family protection”
Customer:
“I want to buy an education
insurance”
Chat Bot:
“We propose the following
options in your situation…”
Chat Bot:
“…and the particular
product parameters are ...”
22. 22Dr. Juergen Rahmel
Understanding Artificial Intelligence Integration
Example: A complex ‘Chat Bot’ solution – a conversational Advisor
Corporate Network
Reasoning
Recognition
recognize
intention
clarify intention
Communication Communication
identify offering
customize
offering
Planning
Prediction
Reasoning
Decision Making
Customer Data Product Data
Customer:
“I am thinking about increasing
my family protection”
Customer:
“I want to buy an education
insurance”
Chat Bot:
“We propose the following
options in your situation…”
Chat Bot:
“…and the particular
product parameters are ...”
23. 23Dr. Juergen Rahmel
Understanding Artificial Intelligence Integration
Example: A complex ‘Chat Bot’ solution – a conversational Advisor
Corporate Network
Reasoning
Recognition
recognize
intention
clarify intention
Communication Communication
identify offering
customize
offering
Planning
Prediction
Reasoning
Decision Making
Customer Data Product Data
Customer:
“I am thinking about increasing
my family protection”
Customer:
“I want to buy an education
insurance”
Chat Bot:
“We propose the following
options in your situation…”
Chat Bot:
“…and the particular
product parameters are ...”
33. Artificial Intelligence in the Insurance Enterprise
David Robson - IBM Watson Group
A one minute introduction to Watson: https://www.youtube.com/watch?v=6SNs9kvRWSA
Modern AIs can ….
Read Natural Language
• News, policies, fact sheets, web sites etc
• Listen and speak
Understand
• understand what it has read or heard and retain this
knowledge at huge scale
Apply Knowledge
• In conversation with people
• Making decisions (medicine, underwriting etc)
Learn with Experience
• Train with experts and during operation
• Improves with experience and feedback
Machine Learning
Deep Learning
Natural Language Processing
34. Common use cases for AI in Insurance
Client Engagement Underwriting Claims management
Client Insight Image recognitionDiscovery
35. Visual Recognition
Analyzes the visual appearance
of images or video frames to
understand what is happening
Language Translator
Translate text from one language
to another
Personality Insights
Understand and engage users on
their own term based on their
personalities and values
Conversation
Hold natural language
conversations with both your
external and internal customers
Speech to Text
Provides highly accurate, low
latency speech recognition
capabilities
Text to Speech
Synthesizes natural-sounding
speech from text
Message Resonance
Communicate with people with a
style and words that suits them
Discovery
Add a cognitive to applications
to identify patterns, trends and
actionable insights
Relationship Extraction
Intelligently finds relationships
between sentences components
(nouns, verbs, subjects, objects)
Tradeoff Analytics
Helps make better choices under
conflicting goals with smart
visualizations & recommendations
Document Conversion
Converts a single HTML, PDF,
or Mic. Word™ document into a
normalized HTML, plain text
A cognitive platform
Tone Analyser
Leverage cognitive analysis to
identify a variety of tones at
sentence or document level
Alchemy Data News
Provides access to an AI
enriched, curated dataset of news
and blog content
DATA
Face Detection/Recognition
Returns the position, age, gender,
and, in the case of celebrities, the
identities of the people in the
photo
Alchemy API
Enable businesses to build apps
that understand the content and
context of text online
36. 2
36
Questions & Answers
How to participate:
If you have a question please type into the messaging area
and send to the presenters
Session format:
The moderator will use a combination of his own questions
and those from the audience
37. Snap Poll3
37
Q. Which of the following use cases for AI / Machine Learning
do you find most compelling
1. Educating consumers about insurance
2. Selling insurance
3. AI as an engagement tool to retain and service customers
4. Managing the claims process and identifying fraud
5. Risk Management & Prevention Advisory services
6. Other
How to participate:
Just respond to the question when it appears on your screen
38. Announcements
Preregistration Open
London 20TH Sept / Singapore 2nd Nov
Following the success of our first Asia conference last year, we will be holding our second Asia annual conference in Singapore
on 2nd November 2017
Our first annual European conference will be held in London on 20th September 2017
Pre-registration for both events is available NOW:
http://asia2017.the-digital-insurer.com/
http://europe2017.the-digital-insurer.com/
Apply For An Award
Applications are now open for Europe and Asia awards
Award categories include the Start-up Insurtech Award and Insurance Innovation Award
Award finalists will present their innovations and solutions at the conferences. The winners will be determined via a live vote
on the conference app from all the attendees
Nominate yourself today via the event website
Entries close on the 5th May for Europe and 26th May for Asia
39. Post webinar activities
Recording will be emailed to registered participants
Next Webinar will be on 17th May 2017 – Digital Transformation Strategies
Register on our website: https://www.the-digital-insurer.com/event/digital-
insurer-webinar-incumbents-fight-back-digital-transformation-strategies/
Please give us your feedback
If you would like to follow up with any of the panelists
- Simon Phipps: simon.phipps@kpmg.com
- Andrew Dart: andrew.dart@the-digital-insurer.com
- Gary Richardson: gary.richardson@kpmg.co.uk
- Adrien Cohen: adrien@tractable.io
- Juergen Rahmel: jr@ietc.hk
- Alberto Chierici: alberto.chierici@spixii.ai
- David Robson: david_robson@uk.ibm.com
Notas del editor
Introducing the panelists
Introducing the discussion agenda
9 world class ML res from leading UK labs
AZ, Peter Dayan, in particular ZG MLG
Work closely w/ Z, proud to have as investor & advisor
30 yrs res experience, 1k citations, expertise lies in DL IL CV
9 world class ML res from leading UK labs
AZ, Peter Dayan, in particular ZG MLG
Work closely w/ Z, proud to have as investor & advisor
30 yrs res experience, 1k citations, expertise lies in DL IL CV
9 world class ML res from leading UK labs
AZ, Peter Dayan, in particular ZG MLG
Work closely w/ Z, proud to have as investor & advisor
30 yrs res experience, 1k citations, expertise lies in DL IL CV