This is the sh@t that undermines the great work we do......
Also, no one said Conversation Design was easy 🤣. But it is satisfying!
Here's my thoughts, anything I've missed❓
✅ You should expect colloquial language ("Y'all") and train for it.
✅ You should also expect short, to-the-point utterances and train for that too.
✅ Instead of vague prompts for repetition, the #chatbot could ask clarifying questions, such as "Could you provide more details about the charge?
✅ The chatbot should be trained to recognise key phrases like "cancelled service" and "representative" to guide the user toward the next steps, such as confirming account details and then routing to a human representative if needed.
✅ Before proceeding with a solution, the chatbot should confirm its understanding of the issue, e.g., "It seems you've been charged post-cancellation, and you'd like to speak with a representative about this, correct?"
Deep Learning Research
5moThe loved the idea just need better work on "non-docs".