Survey chatbots: where are we at, and where are we going? 16 … · 2019. 5. 20. · Survey...
Transcript of Survey chatbots: where are we at, and where are we going? 16 … · 2019. 5. 20. · Survey...
Survey chatbots: where are we at, and where are we going?
16TH May 2019
Simon Neve, CEO
wizu.com
• Simon Neve from Fusion Software
• 18 years creating customer experience solutions
• Creators of Wizu - the first survey chatbot
INTRODUCTION
wizu.com
What We’ll Cover
• Maturity of wider chatbot industry
• Why use survey chatbots?
• Where do chatbots work best / struggle?
• The future of Survey Chatbots?
wizu.com
WIDER CHATBOT MATURITY
# Survey chatbot providers in
– 2017: 1
– 2019: 12*
* https://insightplatforms.com/platforms/category/chatbots-solutions
wizu.com
Commerce
North Face XPS, H & M
eCommerceTransactional
Wizard
Simple Complex
Form
s
Info Delivery
HealthTap
Wizard / NLP
Menu FAQ’sHow-to’sScheduling Agent
Dialogue
Starbucks, RSPCA
NLP / AI
Order TrackingSupport RequestWorkflow ApprovalConcierge
WIDER MARKET BOT TYPES
Info Gathering
DoNotPay
Wizard / NLP
Forms,SurveysLead capture
Virtual Assistant
Alexa, Siri, Cortana
NLP / AI
Message Processing
Hu
man
Mim
icry
Simple Complex
Form
s
SURVEY BOT TYPESH
um
an M
imicry
Forms
Wizard
Quant
Menu basedValidation: simpleData: InboundMaintenance: lowOpen text reporting: Word cloud Bot Skills: noneDesigner skills: DIYCost: $
Validation
Wizard / NLP
Quant + Probing
Menu based with some textValidation: LinguisticData: InboundMaintenance: SomeOpen text reporting: TABot Skills: 20+Designer skills: DIY +Cost: $$
Model Based
NLP / AI / ML
Quant + Probing + Prompting→ QualText/ voice Validation: Linguistic / ModelData: Inbound/ OutboundMaintenance: High / SupervisedOpen text reporting: TA/ MapsBot Skills: 100+Designer skills: Tech / businessCost: $$$
Simple Complex
Form
s
SURVEY BOT TYPESH
um
an M
imicry
Forms
Wizard
Quant
Menu basedValidation: simpleData: InboundMaintenance: lowOpen text reporting: Word cloud Bot Skills: noneDesigner skills: DIYCost: $
Validation
Wizard / NLP
Quant + Probing
Menu based with some textValidation: LinguisticData: InboundMaintenance: SomeOpen text reporting: TABot Skills: 20+Designer skills: DIY +Cost: $$
Model Based
NLP / AI / ML
Quant + Probing + Prompting→ QualText/ voice Validation: Linguistic / ModelData: Inbound/ OutboundMaintenance: High / SupervisedOpen text reporting: TA/ MapsBot Skills: 100+Designer skills: Tech / businessCost: $$$
Simple Complex
Form
s
SURVEY BOT TYPESH
um
an M
imicry
Forms
Wizard
Quant
Menu basedValidation: simpleData: InboundMaintenance: lowOpen text reporting: Word cloud Bot Skills: noneDesigner skills: DIYCost: $
Model Based
NLP / AI / ML
Quant + Probing + Prompting→ QualText/ voice Validation: Linguistic / ModelData: Inbound/ OutboundMaintenance: High / SupervisedOpen text reporting: TA/ MapsBot Skills: 100+Designer skills: Tech / businessCost: $$$
Simple Complex
Form
s
SURVEY BOT TYPESH
um
an M
imicry
Forms
Wizard
Quant
Menu basedValidation: simpleData: InboundMaintenance: lowOpen text reporting: Word cloud Bot Skills: noneDesigner skills: DIYCost: $
Validation
Wizard / NLP
Quant + Probing
Menu based with some textValidation: LinguisticData: InboundMaintenance: SomeOpen text reporting: TABot Skills: 20+Designer skills: DIY +Cost: $$
Model Based
NLP / AI / ML
Quant + Probing + Prompting→ QualText/ voice Validation: Linguistic / ModelData: Inbound/ OutboundMaintenance: High / SupervisedOpen text reporting: TA/ MapsBot Skills: 100+Designer skills: Tech / businessCost: $$$
WIDER CHATBOT MATURITY
wizu.com
wizu.com
wizu.com
Survey Chatbots???
Chatbots
wizu.com
wizu.com
AI PaaS
Conversational AI Platform
WIDER CHATBOT MATURITY
1. Early in days of Chatbot hype
2. Earlier days of Survey Chatbot hype
3. Both based on maturing technology
wizu.com
WHY SURVEY CHATBOTS?
Big problem in the industry:
Survey Fatigue
– Affects completion rates
– Affects quality and quantity of verbatims
– Creates speeding
wizu.com
WHY CHATBOTS? QUALITY
Non-response bias
wizu.com
0
1
2
3
4
5
6
Distribution
WHY CHATBOTS? DEEPER INSIGHT
Case study – Forms vs Chatbot:
• 33% more verbatim text, 118% including AI prompts and probes
• 91% more actionable feedback
wizu.com
WHY CHATBOTS? ENGAGEMENT
Case study – Forms vs Chatbot:
• 57% rated respondent experience 10/10 compared to 11% for forms
• Despite being
1min 35s longer?!
wizu.com
WHY CHATBOTS? ENGAGEMENT
wizu.com
7 7.5 8 8.5 9 9.5
15-24
25-34
35-44
45-54
55-64
65+
How would you rate the experience of taking this survey? (1-10)
Conversational Survey Traditional Survey
SURVEY CHATBOTS – WHERE BEST
• Experience
– Customer / Employee / Patient / Visitor
- Respondent experience is valued
- Known domains
- Reactions based on quality / sentiment
- Millennials / Gen Z
wizu.com
SURVEY CHATBOTS – WHERE NOT
• Opinion polls
• Very long surveys
wizu.com
SURVEY CHATBOTS - CHALLENGES
• Reporting
– Text analytics
• Scripting / Humanization
• Survey fatigue open rates
wizu.com
SURVEY CHATBOTS - FUTURE
• More bot surveys
• More skilful bots
• Blurring lines between surveys and customer service
wizu.com
SURVEY CHATBOTS - FUTURE
- Model driven feedback
- Service Design / Digital Twin
- Model first, then questions
- Provides understanding / context
- Customer Journey Maps perfect for model
wizu.com
CHATBOT EXPERIENCE MODELS
wizu.com
Prepare Arrive Presentations Lunch Leave
• Visit web site
• Read agenda
• Plan journey
• Drop coats at cloakroom
• Get coffee
• Connect to Wi-Fi
• Find seat
• Choose stream
• Eat lunch
• Speak to sponsors
• Pick coat from cloakroom
• Drop lanyardTou
chp
oin
ts
CHATBOT / MODELS
• Taxonomy
wizu.com
Prepare Arrive Presentations Lunch Leave
• Visit web site
• Read agenda
• Plan journey
• Drop coats at cloakroom
• Get coffee
• Connect to Wi-Fi
• Enter main room
• Find seat
• Choose stream
• Choose food from selection
• Speak to sponsors
• Pick up coat from cloakroom
• Drop lanyard
Act
ivit
ies
Intent: “Connect to Wi-Fi”
Entities: • Wi-Fi• Network
Reactions:• Positive: Great ☺• Negative: Right network?
Verbs:• Join• Connect
MODEL ANALYTICS• Heatmaps
Prepare Arrive Presentations Lunch Leave
• Visit web site
• Read agenda
• Plan journey
• Drop coats at cloakroom
• Get coffee
• Connect to Wi-Fi
• Enter main room
• Find seat
• Choose stream
• Eat lunch
• Speak to sponsors
• Pick up coat from cloakroom
• Drop lanyard
Act
ivit
ies
SURVEY CHATBOTS - FUTURE
Mirror human feedback experience
• Close the loop – real-time
• Two way communication
• Express via Voice / Video
wizu.com
SURVEY CHATBOTS - CONCLUSION
• Survey chatbots are coming here
• Offer a solution to the problem of survey fatigue
• Novelty? Or will respondents be even more discriminatory in future?
• Opportunity for survey industry
wizu.com
ANY QUESTIONS?
https://wizu.com
https://linkedin.com/simonneve
Event example: chat.wizu.com/asc/event
wizu.com