2. Getting the customer
job done
Customer experience
in a connected world
Automating conversations to
get the customer job done
Smart. Personal. Easy.
Personal conversations on
every channel or device
15. Digital interaction volume
Growth rate
Live Contact
growth rate?
WE MISSEN DE OPPORTUNITY ONDER ONZE NEUS...
Digital interaction already
10x live
Telematics and IOT will 10x
that
What if 10% is escalated to
live…?
19. 74%
ONDANKS ALLE BIG DATA
Source: ‘Global Contact Centre Benchmarking Report’, Dimension Data, 2015
van consumenten is gefrustreerd met
de onpersoonlijke service online
25. RESULTS
Humans: 90%
Random: 44%
AI High-score:
58%
AI Low-score:
32%
WINOGRAD CHALLENGE
Who is
“he”?
Babar wonders how he
can get new clothing.
Luckily, a very rich old
man who has always
been fond of little
elephants understands
right away that he is
longing for a fine suit.
26. Er is een 50% kans dat
er over 50 jaar sprake is
van HLMI
HLMI = High Level
Machine Intelligence
Source:
WAT DE EXPERTS (WETENSCHAPPERS) DENKEN
27. EEN RETAIL SALES-MEDEWERKER ZOU AL OVER 15 JAAR VERVANGEN KUNNEN
WORDEN DOOR EEN ROBOT. MAAR DE MENINGEN ZIJN VERDEELD
28. Understand customer job (AND GET IT DONE!)
Personalise communication
Be responsive to behaviors & emotions
Reduce uncertainty with notifications
Be helpful and conversational
Deliver on the brand promise
Learn from each conversation
Source:
https://experiencemat
ters.blog/2017/04/12/
report-humanizing-
digital-interactions/ automating conversations
to get the job done
DE ZEVEN BEGINSELEN VAN HUMANISED DIGITAL
ASSISTANCE (& smart chatbots)
Based on
33. • Key objectives:
• Increase online conversion rates from new campaign
• Go live within 10 weeks
• Key assumption:
• Offering automated pro-active, conversational
support in the consumer decision journey can
improve the experience and increase conversion
rates
• Key challenge:
• Offer web chat support with warm transfer, but
prevent contact center to be flooded
Pro-active support in the Consumer Decision Journey
34. • We designed a conversational chatbot that
pops-up pro-actively when the consumer is on
relevant web pages (campaign site, product-
pages, self-service portal etc)
• It is designed to capture 30% to 50% of data that
would normally be captured through a static
form
• Free format data entry is limited to prevent
consumer failure
Use conversations to capture data
35. • The bot captures the name of the pet first and
uses that name in other parts of the conversation
to increase engagement
• We believe that consumers that feel engaged
and have already invested time are more likely to
stick to the end
Make it personal to increase engagement
36. Guiding the customer journey
• The bot compliments the consumer in a natural
way to keep a fluent flow of the conversation
• It guides the consumer to provide answers the
bot is likely to understand to avoid the “sorry I
don’t understand what you are saying”-
experience
37. • At some points the conversation needs to be
escalated to live for underwriting purposes
• At other points in the journey escalation to live has
been designed in with intent to increase the
likelihood for a sale
• We are currently in the process of evaluating based
on data to optimise these escalations
• For the future we are considering to develop a
decision engine with machine learning algorithms to
optimise conversation flows automatically
Intentional escalation to live
38. The consumer chooses the channel
• At the point of escalation to live a warm transfer
to web chat is offered, but the consumer can
make his own choice
• The consumer can also be jumped to the sign-up
form where all known data from the automated
conversation is pre-filled
• Escalation to a live contact occurs in less than 5%
of conversations
39. Make consumers notice a human has stepped in
• The escalation to web chat has been
intentionally designed in a different interface to
mark the transfer from bot to human
• The live agent can review the entire conversation
with the bot and continues where it was
transferred
40. The first results are far above expectations
A/B-tests show that
conversion rates when the bot
is offered are 35% to 100%
higher