46. 1. Call from
customer to
Cisco
UCCE
2. Agent
assigned
call
4. Call is
recorded in
Cisco
MediaSense
3. Agent
gets call
presented
on Finesse
Dashboard
Cognitive Contact
Center Service
5. Call
recording is
collected by
Cognitive CC
Watson Bluemix APIs
Cisco Analytics
• Social Analytics
• Explore/Predict
• Speech to Text
• Natural Language
Classifier
• Alchemy Language
• Tone Analyzer
pipeline
6. Insights
gathered from
Watson APIs
and Cisco
Analytics
Agent reskilling and Line
Manager reporting
7. Insights
Made available
for use for
contact center
operations
1A. Next call
from
customer
7A. Customer
insights
presented on
Finesse
dashboard
Watson Cognitive Contact Center Process Flow
Iterative Process: Insights continually expand and refined with each call for every
customer and every agent
48. Offline Text and Audio Analytics “pipeline”
Expansion of Cloud Pipeline
Meta-data
& Text
Store
Meta-data
& Audio
Store
Meta-data,
text, and
data recording
Analytics
Services
Unstructured
Data
Analysis
Structured
Data
Analysis
Cisco Explore and
Predict
text-
derived
attributes
Audio
Text
Attributes
Text
Product
Attributes
Core
Applicatio
n
Integration
Reskilling
Training
APIs
Admin APIs
Watson APIs
Speech to Text
NLC
Alchemy
Tone Analyzer
Cognitive CC
Store
Agent
Finesse Dashboard
Historic Data provided
during Contact Center Call
Setup
Line Manager
Agent Behavior Reports
Insight Reports
Probable Cause
Time to Close
Future
Stand Alone Micro Service Connector
Automation
Integration
Platform
Future
Reporting APIs
Case
Correlation
Analysis
Future
Next Issue Avoidance
Context Service
Focu
s
Cisco Domain
Specific NLP
Product
Detection
Trainer
UCCE
Data
Store
Insights
Prediction
Classificatio
n
Domain
Meta
Data
Structure
d
Data
Correlation
Predictions
Loop
49. Finesse Widget Integrate into existing Gadgets (e.g. Context Service,
other)
Date- May 2, 2016
Category - Payment not recorded
Overall Sentiment – Positive
Customer Behavior - Extraversion,
Agreeableness Initial Call Tone -
Disgust, Anger
Final Call Tone – Joy
Key Sentence - Per our system’s data, you did
pay your bill last Aug. 12 which was two days ago in one of
our affiliatedpayment centers and you currently have 0
balance
Resolution ?
Date- May 6, 2016
Category - Payment not recorded
Overall Sentiment – Negative
Customer Behavior – Extraversion
Initial Call Tone - Anger
Final Call Tone – Disgust
, Key Sentence - However, the bill that you
received was generated a week before you made the
payment that’s why your latestpayment had not been
reflected
Resolution ?
.
See entire conversation
NLC
Alchemy
Language
Tone Analyzer
Customer
Case Notes
• Pointer into Context Service
• Leverage Context Service to get more info
See entire conversation
TextRank
Representation of Type of
Data: Not Finesse Gadget
Representation
50. Manifest.json
(caller/call metadata)
Caller_Stream_123456.mp4
Agent_Stream_123456.mp4
1. Call happens
2. Call completion triggers recording and
“manifest” generation
“PIPELINE” - a scheduled,
Orchestration service, e.g. Cast
Iron
Process
Manifests
3. Service runs @ scheduled interval
Transcribe
Calls
Persist
Insights
Insights.json
Caller_Stream_Transcript_123456.mp4
Agent_Stream_Transcript_123456.mp4
TTS Dialog-NLC Alchemy Tone
URL’s URL’s
Insights
Operational
Generate
Insights
JSON
51. Watson API Flow
File format : MP4
Subscription : Provides URL
Note: Need to include meta-data with recording
Internal Note: Need to identify which releases
supported
Cisco Recording
Watson Bluemix
Speech to Text
Caller Stream
Agent Stream
Caller TextAgent Text
1
NOTE: Text sent back independently to be used for
other services going forwardSpeech to Text
pipeline
pipeline
Cisco Social
Analytics
52. NLC
Caller Text
NLC
Agent Text
Caller Classifier confidence score
Agent Classifier confidence score
OVERALL CALL Classifier confidence score
2
Alchemy Language
Caller Text
Agent Text
Alchemy Language
• Entities with Relevance , Sentiment and Type
• Keywords with Relevance, Sentiments
• Document Sentiment
• Concepts
• Entities with Relevance , Sentiment and Type
• Keywords with Relevance, Sentiments
• Document Sentiment
• Concepts
• OVERALL CALL
• Entities with Relevance ,
Sentiment and Type
• Keywords with Relevance,
Sentiments
• Document Sentiment
• Concepts
Watson Bluemix
Caller Text Agent Text
• Question: what does Targeted
Sentiment provide above Entity and
Keyword Sentiment ?
Con’d on Next Slide
pipeline
53. Tone Analyzer
Caller Text
Agent Text
2
• Social Summary
• Language Style
• Emotion
• Emotion Trend thru conversation
Watson Bluemix – Con’d
Tone Analyzer
• Social Summary
• Language Style
• Emotion
• Emotion Trend thru conversation
• OVERALL CALL
• Social Summary
• Language Style
• Emotion
Agent Insights
NLC, Alchemy API, Tone Analyzer
OVERALL CALL Insights
NLC, Alchemy API, Tone Analyzer
Caller Insights
NLC, Alchemy API, Tone Analyzer
pipeline
Use this slide but instead of the large “customer journey” bubble, use the “Customer Journey” wrapper from option 3. Have the Customer Journey wrapper be the last build.
Do not have the bubble (voice, email, chat, etc) it move automatically, have them build by click
New ordering for the below:
Reactive
Contextual
Suggestive
Predictive
Move the bubbles on click, do not have it move automatically