3. Why or Why Not Chatbots
• Customer prefer Chatbots
https://www.tidio.com/blog/chatbot-trends-and-stats-in-2019
4. Why or Why Not Chatbots
https://www.tidio.com/blog/chatbot-trends-and-stats-in-2019
5. Why or Why Not Chatbots
• Increasing
sales 7–25%
6. Why or Why Not Chatbots
https://www.comm100.com/blog/how-much-can-chatbots-actually-save-you.html
Save up to $0.7
per interaction
Cost savings of
$20 million globally
$8 billion per year
by 2022
• Saving cost
8. Why or Why Not Chatbots
• Cost reduction
• Reducing response time
• Increasing customer convenience
• Easy, Fast and Natural
• Increasing first time resolution
15. • Skills Chatbots
– specific task driven
– single-turn-type bots
– less contextual awareness
– sometimes command based
– prefer speech to text
Types of Chatbots
Image from https://voicebot.ai/amazon-echo-alexa-stats/
17. • Support/Service Chatbots
– to master a single domain
– to answer a wide range of FAQ-type
questions
– to engage the audience
– to give recommendations
– to provide on-demand services
– to have contextual awareness
Types of Chatbots
https://www.invespcro.com/blog/chatbots-customer-service/
18. What Are They Good At
• Focused Scope
• High volume of requests
• Problems of routine nature
• Multiple steps or input parameters
• Peace of mind
20. How do Chatbots Work
• Rule based Approach
• Utterance and Response pairs are pre-defined in graphed architecture
• Expensive human labour to build hand-craft rules
• Templates are provided to minimize the coding effort
21. Rule based Approach
• Pandorabots Platform
• Built with AIML Engine
• Useful for extremely narrow domain
23. Retrieval-Based Approach
• Chatterbot
• Python library for building Chatbots
• Conversation data collection and Python Programing skills are needed
• Easy to build and iterate
• Widely used in FAQ and Website Chatbots
24. NLP Driven Approach
• NLP Driven Approach
• Natural Language Processing:how to program computers to
process and analyze large amounts of natural language data.
• Using ML/DL, NLP is able to analyse unseen data, especially when a
(huge) volume of example (training) data is available.
http://customerthink.com/5-reasons-why-your-chatbot-
needs-natural-language-processing/
Deep
Learning
N
L
P
Machine
Learning
Chat
bot
AI
32. NLP In Chatbots
• To understand the request:
• What is the weather in Seattle today?
• What’s the weather in London?
• Tell me the temperature now.
Question-word
Command-word
Intent-word
Location-word
Time-word
WeatherForecast
Seattle / London
Today / Now
.
Intent detection
Slots detection
sample utterances
33. Why Chatbots Need NLP
• Ability to understand
complicated/unseen utterances
“ I’ve been a loyal customer
for 8 years and your product
won’t let me upload a second
W-2. Can someone please
help me with this!?”
34. Platforms to Build Chatbots
Self-service solutions
with GUI
(Drag and Drop)
Self-service solutions
with an API, SDK or
library
End-to-end
solutions
Speed Fast Medium Fast
Cost Low to Medium Medium Medium to High
Flexibility Low High Medium to High
Coding Effort
Required
None to Low Medium to High None
Technique
behind
Rule + Retrieval
Some NLP
Mostly NLP driven Hybrid
Platform ChatBot
Chatfuel
ConversationBuilder
DialogFlow
IBM Watson
Alexa
DialogFlow
Azure Bot Service
wit.ai
Chatterbot
Intercom
Conversica
Drift
Nuance
ubisend
https://aimultiple.com/chatbot-platform
35. Dos and Don’ts
• Service within narrow use-cases
• Ask simple questions with options
provided
• Acknowledge the AI identity
• Chatbots sit on the driving seat
• Handover to human agent as soon as
needed seamlessly
• Keep upgrading for new scenarios
• Try to do too much
• Ask too many questions and expect
complicated answers
• Challenge the Turing Test
• Ask open questions
• Leave the Chatbot alone
• Deal with complaints
• Save the maintenance effort
36. Cook Chatbots UP
• Build FAQ Chatbots based on DialogAI
• Google account
• Integrate Chatbots to Facebook Messenger
• Facebook account and Facebook page for Travel
• Knowledge source (FAQ) about travel
• https://www.onetravel.com/faq/
• https://www.intrepidtravel.com/en/contact-us/frequently-asked-questions
37. Closing Thoughts
• Will I lose control of the customer experience?
• Which type of robots should I build?
• How to form a Chatbot project team?
• Are Chatbots cutting jobs?
40. Reference
• Galitsky, Boris. "Developing Enterprise Chatbots." (2019).
• McTear, Michael, Zoraida Callejas, and David Griol. The conversational interface: Talking to
smart devices. Springer, 2016.
• Xu, Anbang, et al. "A new chatbot for customer service on social media." Proceedings of the
2017 CHI Conference on Human Factors in Computing Systems. ACM, 2017.
• Brandtzaeg, Petter Bae, and Asbjørn Følstad. "Why people use chatbots." International
Conference on Internet Science. Springer, Cham, 2017.
• https://blog.aimultiple.com
• https://www.invespcro.com/blog/chatbots-customer-service
• https://chatbotsmagazine.com
• http://customerthink.com/5-reasons-why-your-chatbot-needs-natural-language-processing/
• Live Chat Benchmark Report 2019 by Comm100