The lifecycle of a
chatbot
Sohan Maheshwar
• Leads Developer Relations at Gupshup
• @soganmageshwar
• Likes comic books and frisbees
“The difficulty in building a chatbot is less a technical one and more an issue
of user experience. The most successful bots will be the ones that users want
to come back to regularly and that provide consistent value.”
- Matt Hartman, Director of Seed Investments at Betaworks
• Software, websites and apps have a dev. lifecycle
• Bot functionality is only going to get more complex
• Formal framework around the bot lifecycle
Requirement
• Gather market requirements
• Create user stories with
conversation in mind
Spec
• Identify the features and
functionality of the bot
• Choose your messaging
platform(s)
• NLP usage
Script
• The conversation is the interface
• Getting input from the user
• Variables in input
• Slot Filling
• Hand-crafting output
• Character
• Friendly
• Guide the user
• Script writers
Architect
• Engineering design for the
bot
• Data persistence -
Personalisation & Context
• Cross platform APIs
Develop
• Range of tools for bot
development
• Ready-to-publish templates
• Conversation editors
• Cloud-based IDEs
• Custom development
• Simplicity is control
Test
• Diversity in messaging apps
• Difference in rendering
messages
• Bot testing tools
Deploy
• Deployed to a hosted
environment.
• The hosted environment must
be stable & needs its own
monitoring and devops support
Publish
• Approval process for each
messaging platform
• Approval time ranges from
instant (Telegram) to a few days
(Facebook)
• Submissions require bot assets
(icons, videos)
Monitor
• Monitor using actual
conversational scripts.
• The bot may be unresponsive
to certain user conversations.
Promote
• Bot Store Optimisation
• Cross-platform bot stores
• Third party referrals
Analyze
• Messages/session
• Retention
• Location, gender
demographics
• Sentiment analysis
Requirement
• Infer new requirements from
analytics
• Create new user stories
Questions?
© Copyright Gupshup 2016
Gupshup has been a leader in messaging and bots
for many years
19
Cloud APIs for SMS
• Leading cloud
messaging platform
handling 3-4 billion
messages per month.
Launched 2010.
Teamchat messaging
app
• Pioneered structured
messages in chat. Built
and deployed
hundreds of bots.
Launched 2014.
Bot Platform
• Leverages prior
experience and
expertise. Since f8,
2500 bots created,
paying customers.
Launched 2016.
Gupshup has been enabling
advanced messaging-based
services (now called bots)
globally since 2010.
Happy Bot Building!

The lifecycle of a chatbot

  • 1.
  • 2.
    Sohan Maheshwar • LeadsDeveloper Relations at Gupshup • @soganmageshwar • Likes comic books and frisbees
  • 3.
    “The difficulty inbuilding a chatbot is less a technical one and more an issue of user experience. The most successful bots will be the ones that users want to come back to regularly and that provide consistent value.” - Matt Hartman, Director of Seed Investments at Betaworks
  • 4.
    • Software, websitesand apps have a dev. lifecycle • Bot functionality is only going to get more complex • Formal framework around the bot lifecycle
  • 6.
    Requirement • Gather marketrequirements • Create user stories with conversation in mind
  • 7.
    Spec • Identify thefeatures and functionality of the bot • Choose your messaging platform(s) • NLP usage
  • 8.
    Script • The conversationis the interface • Getting input from the user • Variables in input • Slot Filling • Hand-crafting output • Character • Friendly • Guide the user • Script writers
  • 9.
    Architect • Engineering designfor the bot • Data persistence - Personalisation & Context • Cross platform APIs
  • 10.
    Develop • Range oftools for bot development • Ready-to-publish templates • Conversation editors • Cloud-based IDEs • Custom development • Simplicity is control
  • 11.
    Test • Diversity inmessaging apps • Difference in rendering messages • Bot testing tools
  • 12.
    Deploy • Deployed toa hosted environment. • The hosted environment must be stable & needs its own monitoring and devops support
  • 13.
    Publish • Approval processfor each messaging platform • Approval time ranges from instant (Telegram) to a few days (Facebook) • Submissions require bot assets (icons, videos)
  • 14.
    Monitor • Monitor usingactual conversational scripts. • The bot may be unresponsive to certain user conversations.
  • 15.
    Promote • Bot StoreOptimisation • Cross-platform bot stores • Third party referrals
  • 16.
    Analyze • Messages/session • Retention •Location, gender demographics • Sentiment analysis
  • 17.
    Requirement • Infer newrequirements from analytics • Create new user stories
  • 18.
  • 19.
    © Copyright Gupshup2016 Gupshup has been a leader in messaging and bots for many years 19 Cloud APIs for SMS • Leading cloud messaging platform handling 3-4 billion messages per month. Launched 2010. Teamchat messaging app • Pioneered structured messages in chat. Built and deployed hundreds of bots. Launched 2014. Bot Platform • Leverages prior experience and expertise. Since f8, 2500 bots created, paying customers. Launched 2016. Gupshup has been enabling advanced messaging-based services (now called bots) globally since 2010.
  • 20.