Looking Back on Three Years Building Bots
summary
EDF Energy built an Alexa skill at the UK Echo launch (2016) and then developed voice and Facebook Messenger chatbots, learning that tone, platform constraints, confirmation and discoverability are crucial for customer interactions—customers expect a serious tone for billing. In 2018 the team created a collective‑intelligence bot combining a graph database, recommender system and NLP to produce personalized answers, plus a multimodal voice+web prototype over WebSockets, both of which proved useful and progressed toward customer rollout.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 3:16Dialogflow acquisition
- 5:51Initial chatbot mistakes
- 9:11Text chatbot language
- 10:56Voice system limits
- 14:26Collective intelligence
- 15:55Voice and visual multimodality
- 17:20Bots versus AI
- 19:04Platform and UX
- 20:47Bot discoverability
- 22:54When bots fail
- 25:47Q&A
related talks
42mLet’s Chat-Bots using Azure AI
27mLet's Chat-Bots using Azure AI
55mBuilding for Alexa with Web API
44mDelivering On The Promise Of Voice With Alexa
54mChatbots: The hype, challenges and opportunities
56mGoogle vs Alexa: Battle of the Bots
52mThe most natural UI. Building voice-enabled experiences with Amazon Alexa
54mTalk over text: Conversational interface design and usability