How not to fuck up your bot UX
summary
Demonstrates building an ecommerce Facebook Messenger bot using Botpress with Google-based caching, prototyping conversation flows via live roleplay, and measuring success with metrics like conversations-per-session and conversion. Explains common failure modes (deadlocks, server lag, lost state) that caused up to 73% dropoff after a bad first contact, gives historical successes (e.g., 1.5 million invites in a Chinese chat experiment), and recommends modular use-case trees, quick-reply UI, short texts, logging and caching, plus human handover for outages.
AI-written summary · about
chapters (37)
- 0:00Introduction
- 2:50Bots and AI history
- 9:22Tech milestones
- 12:14Market risks
- 13:23Bad UX examples
- 19:18Channel mistakes
- 20:01Bot goals
- 22:08Acceptance criteria
- 23:51Brand expectations
- 24:35Platform features
- 26:20UI interactions
- 27:47Prototyping feedback
- 29:53Digital nativity
- 30:57Modular architecture
- 31:39Natural dialogue
- 32:44Integration validation
- 33:04Metrics behavior
- 34:08Language emotions
- 34:49Game loop
- 36:54Error escalation
- 37:57Feedback improvement
- 38:40Error communication
- 39:42Interface design
- 40:48Caching load times
- 41:29Team roles
- 42:10Maintenance costs
- 42:54Human escalation
- 43:38Q&A
- 45:10A/B testing multilingual
- 46:37Live bots limits
- 47:18Loyalty use case
- 48:02Profitability
- 49:06Quick replies FAQ
- 50:42Dialect challenges
- 51:44Commerce success
- 52:47Technical frameworks
- 53:33Summary
related talks
50mThe Art of Coding a Conversation: Designing Bots
27mA primer to write a conversational chatbot
54mBuilding advanced ChatBots
35mChat Bot Building Story - The Good, the Bad and the Unexpected
54mChatbots: The hype, challenges and opportunities
13mWhy Chatbots Fail, And How To Fix Them
55mEnterprise Chatbots (using messaging and AI to build novel user interfaces for work)
55mConversations for the masses