3. Conversational AI for the future of business
Notify Engage Empower Support
Iris warms up your audience and customers so that your team can have the
conversations that matter.
4. Intent clashes + Selective confidence
● What to do when you hit intent clashes?
● More data will only get you so far
● The “refund” problem, want one or just information about it?
● How do you achieve accuracy and consistency?
● How can this be solved through selective confidence?
5. Dialogue Design
● We have always called it Dialogue Design but essentially it’s CDD
● It’s an approach / methodology / practice for success
● At the core it’s customer centric development
● Human-in-the-loop Conversational product development
6. Building for CX
● Thousands of customer interactions
● Three customer segments
● Organic and paid traffic
● Integration into their existing systems
● Complex business rules
● Over 10 different scenarios
8. CDD by design and practice
Dialogue Designer
● Rapid prototyping
● Copy editor
● Low level programming
● Integration interfaces
● Conversation flow
● Encode business logic
Launch Pad / Insights & Metrics
● Design
● Staging
● Production
● Review
● Assess
● Rapid development workflow
Conversation Engine
● Various services + functionality
● Rasa service
● Rasa for Traversal or Article Search
● Various training models
● Processed at different levels
● User profile, settings and context
Natural Language Interface
● Intent training
● Text input and keyboards
● Set the confidence thresholds
● Simulation + load testing
● Capture failed NLU
● Retrain or discard
10. Key takeaways
● Select intents based on context to reduce clashes
● Leverage context to adjust confidence threshold
● How can you tighten the CDD workflow
● Set an order of operations on your NLU