Chat Bot Building Story - The Good, the Bad and the Unexpected
summary
A bank-grade youth chatbot called Caching goes live after months of iteration and rolls from beta to production in April 2017, trained on 120 intents and scaled across multilingual pipelines using LUIS and IBM Watson, with about 40 bots in test at peak. Unexpected real queries force features: a suicide hotline redirect, emoji normalization, adaptive cards for date input, and rapid DevOps releases (sometimes 20 deploys a day); outages once trigger costly helpdesk traffic.
AI-written summary · about
chapters (23)
- 0:00Intro
- 0:15Anxiety
- 0:30About me
- 1:11How it all started
- 2:25The Hype Cycle
- 3:59Forbes Bots 20
- 5:01Lifecycle
- 5:36Brain Mapping
- 7:10Caching
- 8:32HR
- 9:00User flows
- 9:50Unexpected things
- 11:16Suicide
- 13:48Jokes
- 16:23Advice
- 18:07Life Cycles
- 21:09NLP Engine
- 22:02UX
- 26:59Adaptive Cards
- 27:43Visual Designer
- 30:11Templating
- 32:37Microsoft Technology
- 34:48Outro
related talks
54mIntroduction to Chatbots
36mChatbots Past, Present and Future
56mBots - the next UI revolution
35mChatbots: The hype, challenges and opportunities
35mDesigning Experiences for Chatbots
54mChatbots: The hype, challenges and opportunities
50mThe Art of Coding a Conversation: Designing Bots
25mFantastic chatbots and where to find them