Build an AI based
Virtual Agent
Ashish
Twitter: @ashish_fagna
Chatbot EveryWhere !
Chatbot to Voice Assistants
Gartner Observations :
“Predicts 2017 - Artificial
Intelligence”
By 2019, more than 10% of IT hires in customer service (ITSM) will write scripts for
bot interactions.
Organisations using Artificial Intelligence based systems would achieve four time
more often than others. (4X)
By 2020, 20% of companies would dedicate workers to monitor and guide neural
networks.
By 2019, startups would overtake Amazon, Google, IBM and Microsoft in driving the
artificial intelligence economy with disruptive business solutions.
By 2019, artificial intelligence platform services will cannibalise revenues for 30% of
market-leading companies
Source : http://www.gartner.com/imagesrv/media-
products/pdf/rage_frameworks/rage-frameworks-1-34JHQ0K.pdf
Technology Trend : Virtual
Assistants
source: https://www.forbes.com/sites/gartnergroup/2017/08/18/future-trends-in-the-gartner-hype-cycle-for-emerging-technologies-2017/#7a5d20214b97
https://www.gminsights.com/industry-analysis/intelligent-virtual-assistant-iva-market
Virtual Assistants Market
Size
Applications in Domains
source: http://press.trendforce.com/node/view/2738.html
Virtual Assistants Market
Size
• Amazon Echo has 70% of Virtual Personal Assistant (VPA) Market.
• Google home has 30% VPA Market.
• Gartner suggests that VPA-enabled wireless speakers will generate $3.52BN in
global revenue by 2021, up from $0.72BN in 2016.
source: https://techcrunch.com/2017/08/25/putting-the-voice-assistant-speaker-craze-in-context/
Amazon Skills Market
Size
source: https://www.voicebot.ai/2017/07/02/amazon-alexa-skill-count-passes-15000-in-the-u-s/
#MachineLearning
#ArtificialIntelligence
Machine Learning Basics
Neural Networks• Neural networks, is a programming paradigm which enables a computer to learn from observational data
• Deep learning, a powerful set of techniques for learning in neural networks
• Neural networks and deep learning currently provide the best solutions to many problems in image
recognition, speech recognition, and natural language processing.
Alexa Use Cases
Alexa Skills Kit (ASK)
• Skills are the New App.
Alexa : Terminology
Alexa, ask <<Invocation_Name>>, Whats the news ?
Alexa, launch <<Invocation_Name>>, <<Intent>>
Add New Skill
Architecture : Alexa Skill
Alexa Skill has two parts:
1. Configuration Data in Amazon Skill Developer Console
(https://developer.amazon.com/edw/home.html#/skill)
2. Hosted Service hosted either on AWS Lambda or any
other HTTPS secured web server (NodeJS)
https://aws.amazon.com/lambda/
Skill Interaction Model
Alexa Skill Architecture
Alexa Web based
Simulator
Demo Use-Case #1:
Get Ticket Details from Service Now.
“Alexa, ask Virtual Agent to tell me about incident number
INC0010315”
ASK : Request Process
ITSM Tools
• ServiceNow has succeeded by
modernising the IT service
management (ITSM) market
through an innovative platform
that is cloud-based saas
platform and highly flexible.
• Valuation more than $10B.
ITSM Automation
Benefits
source: https://servicematters.servicenow.com/2017/08/07/one-step-better-customer-service/
• Reduced need to depend on people, as
best practices are clearly defined and
ITSM automation is leverages
technology to do the heavy lifting
• Less people dependency brings greater
cost savings
• Significant reductions in human error
• Accelerated and improved incident
response process
• Greater IT service delivery process,
which improves both internal service
levels and external customer
Demo Use-Case #2:
Connect To Wi-Fi ( after mobile PIN
verification).
“Alexa, ask Virtual Agent to tell me how to connect to wi fi as a guest
user”
“Alexa, tell Virtual Agent My Mobile Number is +XX XXXX XXX XXX”
“Alexa, tell Virtual Agent PIN Received is XXXX”
Neva.ai : Makes Human Agent Smarter !
• Neva developed an AI based decision engine that sits on
ServiceNow.
• Applying machine learning and natural language processing to
automate customer service and support.
• Is able to interact conversationally with requestors to answer any
questions related to their self service issues.
• Harness the power of historic ITSM data to answers customer
requests faster.
Neva.ai : Makes Human Agent Smarter !
• Example 1: HR based Use Case, Neva makes context
based recommendations that makes fulfiller accelerate the
resolution of that case.
• Example 2: Neva can also be used to predict values of
certain fields for incoming new incident.
Neva + ServiceNow
Source: https://www.youtube.com/watch?v=BPUtufb7Hz0
Fulfiller Assistance
source: https://www.slideshare.net/danturchin/the-future-of-ai-in-it-dan-turchin-sfhdi-march-2017
Demo Use-Case #3:
Send Email to ServiceNow (Create new
incident, neva.ai auto-maps this incident
using AI & ML)
“Alexa, ask Virtual Agent to send email to ServiceNow. The subject
is laptop does not start and the body is I upgraded my laptop”
How Neva Makes Predictions
?
• Neva machine learning models use supervised learning
and a variety of algorithms to build ML models from
historical data (incidents, user ticket history, user
geography, etc.)
• Neva uses a neural net but compare prediction accuracy
across several algorithms simultaneously and use the
best-performing one.
Thank You
Ashish
Twitter: @ashish_fagna

Build an AI based virtual agent

  • 1.
    Build an AIbased Virtual Agent Ashish Twitter: @ashish_fagna
  • 2.
  • 3.
    Chatbot to VoiceAssistants
  • 4.
    Gartner Observations : “Predicts2017 - Artificial Intelligence” By 2019, more than 10% of IT hires in customer service (ITSM) will write scripts for bot interactions. Organisations using Artificial Intelligence based systems would achieve four time more often than others. (4X) By 2020, 20% of companies would dedicate workers to monitor and guide neural networks. By 2019, startups would overtake Amazon, Google, IBM and Microsoft in driving the artificial intelligence economy with disruptive business solutions. By 2019, artificial intelligence platform services will cannibalise revenues for 30% of market-leading companies Source : http://www.gartner.com/imagesrv/media- products/pdf/rage_frameworks/rage-frameworks-1-34JHQ0K.pdf
  • 5.
    Technology Trend :Virtual Assistants source: https://www.forbes.com/sites/gartnergroup/2017/08/18/future-trends-in-the-gartner-hype-cycle-for-emerging-technologies-2017/#7a5d20214b97
  • 6.
  • 7.
    Virtual Assistants Market Size •Amazon Echo has 70% of Virtual Personal Assistant (VPA) Market. • Google home has 30% VPA Market. • Gartner suggests that VPA-enabled wireless speakers will generate $3.52BN in global revenue by 2021, up from $0.72BN in 2016. source: https://techcrunch.com/2017/08/25/putting-the-voice-assistant-speaker-craze-in-context/
  • 8.
    Amazon Skills Market Size source:https://www.voicebot.ai/2017/07/02/amazon-alexa-skill-count-passes-15000-in-the-u-s/
  • 9.
  • 10.
  • 11.
    Neural Networks• Neuralnetworks, is a programming paradigm which enables a computer to learn from observational data • Deep learning, a powerful set of techniques for learning in neural networks • Neural networks and deep learning currently provide the best solutions to many problems in image recognition, speech recognition, and natural language processing.
  • 12.
  • 13.
    Alexa Skills Kit(ASK) • Skills are the New App.
  • 14.
    Alexa : Terminology Alexa,ask <<Invocation_Name>>, Whats the news ? Alexa, launch <<Invocation_Name>>, <<Intent>>
  • 15.
  • 16.
    Architecture : AlexaSkill Alexa Skill has two parts: 1. Configuration Data in Amazon Skill Developer Console (https://developer.amazon.com/edw/home.html#/skill) 2. Hosted Service hosted either on AWS Lambda or any other HTTPS secured web server (NodeJS) https://aws.amazon.com/lambda/
  • 17.
  • 18.
  • 19.
  • 20.
    Demo Use-Case #1: GetTicket Details from Service Now. “Alexa, ask Virtual Agent to tell me about incident number INC0010315”
  • 21.
  • 22.
    ITSM Tools • ServiceNowhas succeeded by modernising the IT service management (ITSM) market through an innovative platform that is cloud-based saas platform and highly flexible. • Valuation more than $10B.
  • 23.
    ITSM Automation Benefits source: https://servicematters.servicenow.com/2017/08/07/one-step-better-customer-service/ •Reduced need to depend on people, as best practices are clearly defined and ITSM automation is leverages technology to do the heavy lifting • Less people dependency brings greater cost savings • Significant reductions in human error • Accelerated and improved incident response process • Greater IT service delivery process, which improves both internal service levels and external customer
  • 24.
    Demo Use-Case #2: ConnectTo Wi-Fi ( after mobile PIN verification). “Alexa, ask Virtual Agent to tell me how to connect to wi fi as a guest user” “Alexa, tell Virtual Agent My Mobile Number is +XX XXXX XXX XXX” “Alexa, tell Virtual Agent PIN Received is XXXX”
  • 25.
    Neva.ai : MakesHuman Agent Smarter ! • Neva developed an AI based decision engine that sits on ServiceNow. • Applying machine learning and natural language processing to automate customer service and support. • Is able to interact conversationally with requestors to answer any questions related to their self service issues. • Harness the power of historic ITSM data to answers customer requests faster.
  • 26.
    Neva.ai : MakesHuman Agent Smarter ! • Example 1: HR based Use Case, Neva makes context based recommendations that makes fulfiller accelerate the resolution of that case. • Example 2: Neva can also be used to predict values of certain fields for incoming new incident.
  • 27.
    Neva + ServiceNow Source:https://www.youtube.com/watch?v=BPUtufb7Hz0
  • 28.
  • 29.
    Demo Use-Case #3: SendEmail to ServiceNow (Create new incident, neva.ai auto-maps this incident using AI & ML) “Alexa, ask Virtual Agent to send email to ServiceNow. The subject is laptop does not start and the body is I upgraded my laptop”
  • 30.
    How Neva MakesPredictions ? • Neva machine learning models use supervised learning and a variety of algorithms to build ML models from historical data (incidents, user ticket history, user geography, etc.) • Neva uses a neural net but compare prediction accuracy across several algorithms simultaneously and use the best-performing one.
  • 31.