Using artificial intelligence (AI) to solve real work problems is no longer science fiction. AI is something that is becoming more common in our daily lives, and it has the potential to enhance the way people use technology in their daily activities. The University of Manitoba is just one such enterprise that is already looking into how AI can improve the student experience and aid our staff in doing their day to day tasks.
> But where do we start?
> In this session, we’ll start with a primer on the types of AI available today that can be used to build modern applications the enhance the user experience for our clients. Then, we’ll take a deep dive by following up with a post-mortem on IST’s own chatbot initiative to explore how the University of Manitoba could use chatbots to provide a robust scalable solution for helping students.
Delivered at Prarie Dev Con in Winnpeg, MB, Canada on June 5th, 2018.
7. What is this talk?
• AI
• The Enterprise
• And You
8. AI in the Enterprise
Scenario:
Your enterprise manages hundreds of support tickets day that need to
be routed to the appropriate support team.
Create a utility that provides this functionality.
9. You
• AI apps are trained to understand rather than programmed to
understand
• AI enhances our work tools, rather that being our work tools
15. • Vocabulary for your chatbot
Entities
• Action the user wants to perform
Intents
• Sample dialog used to train the chatbot
Utterances
• The map of the conversation
Dialog Tree
16. Send a message to David Wesst
Entity
@member: David Wesst
Intent
#send_message
17. Drop a line to DW
Entity
@member: David Wesst
Intent
#send_message
20. LUIS
• Microsoft
• https://luis.ai
• Highlights
• Improved UI
• Large number of pre-trained sets
• Multiple pre-built integrations
with Azure services
21. Watson Assistant
• IBM
• https://www.ibm.com/watson/
• Highlights
• Most Mature
• Industry focused
• Integrates with BotKit and Bot
Framework
26. How do you this? (Host App)
• BotKit
• Microsoft Bot Framework
27. How do you this? (Host App)
BotKit
• Open Source
• Tool to integrate directly with
some platforms
• BotKit Studio aids with AI
development
• Hosted Development Environment
• Integrates with MSFT Bot
Framework
Bot Framework
• Largest number of native
integrations
• Includes Azure Integrations
• Pay for “premium” connections
• Not an actual framework, rather
a platform
28.
29. How do you this? (Host App)
Microsoft Bot Framework BotKit
32. UM Student Chatbot
• Proof of Technology Project
• Determine if the university can use AI to provide business value to clients
• Solution needs to provide business real value direct to clients
• Integrate with existing client facing systems
34. How did your team do this?
• Coaching & Support
• Defining entities, intents, and using the Watson development environment
• Controlling scope
• Software Development 101
• Bot Testing
• Built the Host
What is this talk AI Landscape
How do we do this AI Development Primer
What could go wrong Post-Mortem / Lessons Learned
Hard problems when you need to define a finite set of business rules and limited amount of sample data
Transition – So what sort of AI is available for enterprise solutions?
Trained AI – You pay for access to AI trained on large datasets
Conversational AI – NLP communication, which combines different trained AI to derive user intent
Custom AI – You train it on your data
Part 1 -- AI
Photo Credit: https://pixabay.com/photo-2356433/
Vendors provide similar offering and work in a similar way
Hosted AI Development Environment
Photo by Amy Skyer on Unsplash