Module 2

How to Measure the Success of

Your IVR System and Make
Data-Driven Improvements
Session 1
 Agenda

1 What is IVR and it’s Components

2 IVR Process Flow

3 Backend Algorithm and Logic

4 Key Metrics to Measure IVR Success

5 Data-Driven Improvements for IVR

Systems
6 Building a Roadmap for Continuous
IVR Improvement
7 Case Study & Role Playing Simulation

2
Quick Recap

6
 Technology Evolution: IVR

Stage 1 Stage 2 Stage 3 Stage 4

Foundational Systems Enhanced IVR Systems Intelligent IVR Systems

Omnichannel & Conversational AI
1980s-1990s 1990s-2000s 2000s-2010s (2010s-Present)

• Automated telephone • Sophisticated voice • Adoption of AI & Machine • Seamless integration across various
systems for basic call recognition capabilities & Learning for more intuitive communication channels, providing a unified
routing and
Description

Natural Language Processing interactions customer experience.

information retrieval • Enhanced menu and user- • Advance analytics to track and • Use of conversational AI and chatbots to enable
• Simple touch tone friendly interactions improve user experiences and natural language interactions via voice.
menu • Integration with back-end outcomes • Enhanced focus on customer engagement and
• Limited functionality system for personalized • Integration with multi channel satisfaction through personalized experiences.
to navigate experience systems

• Call routing to • Personalized greetings based • Contextual assistance based on • Start interactions on one channel and continue
Use Cases

department on caller ID previous interactions and data on another (e.g., from voice to chat).
• Simple information • Complex transaction handling • Predictive routing to direct calls • AI-driven conversational IVR for more natural
retrieval • Call-back options for high- • Real time feedback & and fluid dialogues with customers.
• Automation for high volume periods to reduce wait adjustments based on caller • Proactive outreach based on customer data
volume queries times behavior and behavior patterns.

Nortel Avaya Genesys Cisco Amazon Connect Twilio

Nuance
Five 9
 Components of IVR Systems
MercuryTelephony Interface:
Servers: Media Playback Devices:
• Connects the IVR system to the
• Host the IVR application and manage • Equipment that plays pre-
telephone network (PSTN or VoIP).
call processing. recorded messages and prompts
Hardware • Handles call signaling and media
• May include dedicated servers or to callers.
Components processing. is the closest planet to
cloud-based solutions. • Ensures high-quality audio output
the Sun

IVR Software: Database Management System: Speech Recognition Software:

• Enables natural language
• Defines the call flows, prompts, and user • Stores customer data, call logs, and processing for understanding
interactions. IVR configurations. voice commands.
Software • Often includes features for reporting and • Ensures data accessibility for • Improves user experience by
Components analytics. personalized interactions. allowing conversational
interactions.

Call Flow Scripts: Prompts:

• Detailed logic outlining how calls are • Pre-recorded messages guiding callers
processed. through the system.
User Interface • Defines the sequence of prompts based • Should be clear, concise, and easy to
Components on user input. understand.
 IVR Process Flow

01 02
Initial Call Reception User Interaction

• Calls are routed through the telephony interface • Callers navigate the IVR system using DTMF (Dual-
to the IVR system. Tone Multi-Frequency) inputs or voice commands.
• The system greets the caller with a welcoming • The system recognizes user input and directs the
message. flow accordingly.

04 03
Call Termination Example Work Flow

• Calls conclude based on user needs—either

transferring to an agent or providing information via • Greeting: “Welcome to [Company Name].”
the IVR. • Main Menu: “Press 1 for Billing, Press 2 for Technical Support.”
• Offers options for feedback or survey post- • Submenus: Each selection leads to additional options (e.g., “Press 1 for
interaction. account information”).
• Data Collection: Collect necessary information (e.g., account number,
reason for call).
• Final Resolution: Directs the caller to an agent or provides automated
solutions.
 Backend Algorithm and Logic
Algorithm Overview:

Algorithms dictate how the IVR system processes calls and makes decisions based on user
input.

Key Algorithm Components

Input Processing Condition Checks Response Generation Data Retrieval

• Evaluates user input
• Captures DTMF tones against predefined • Creates responses
• Accesses relevant
or voice commands logic. based on user input
customer data from
from the user. • Determines the next and system logic.
the database for
• Converts inputs into step in the call flow • Ensures responses
personalized
actionable data for based on conditions are accurate and
responses.
the system. (e.g., if a user presses timely.
1)

11
 Conversational IVR
Overview
Conversational IVR goes beyond traditional menu-driven IVRs by leveraging Natural Language Processing (NLP) to understand
and respond to spoken language naturally. This eliminates the need for users to navigate complex menu trees.
Features

Natural Language Understanding (NLU): Contextual Awareness: Real-time Interaction:

Key

Recognizes customer intent based on free- Uses historical interactions to provide relevant Customers can speak their requests
form speech options and personalized responses. without predefined options.
Benefits

Improved Customer Experience: Increased Containment Rate: Enhanced Personalization:

Faster issue resolution by minimizing Higher self-service completion with Tailors responses based on customer
frustration with menus conversational flows. data and previous interactions

Telecom Example: A conversational IVR allows customers to state their issues, like “I need to reset my internet,” which the system
immediately handles without menu options.

Containment Rate: Abandonment Rate: Customer Effort Score (CES):

Key Metrics % of calls dropped during the IVR flow.
% of customers completing interactions How easy customers find it to resolve
through self-service. their issue via IVR.
 Conversational IVR: AI-Driven Customer Journey
Mapping
Conversational IVR is not just about replacing traditional IVR systems; it's about creating personalized, data-driven customer journeys. AI in
Conversational IVR can dynamically adapt to customer needs, improving both experience and operational outcomes.

Context-Aware Routing: Conversational IVR can AI-Powered Personalization: By leveraging customer data
dynamically change the call flow based on the customer's (from CRM, past interactions), the IVR can personalize each
interaction history and current query. conversation.
Advanced Features
Example: A returning customer with a service issue is Example: Instead of asking repetitive questions, the IVR
routed directly to the support team without navigating a uses stored preferences, reducing the interaction time by
menu. 30%.

Advanced Use Cases Impact on Call Center KPIs

• IVR Containment Rate: AI-driven
Proactive Support: AI-powered IVR Multilingual Support: Conversational IVR personalization can improve
systems can offer proactive solutions by systems can switch languages based on containment rates by up to 35%.
predicting common issues. For example, customer preferences, offering a seamless
telecom customers can be notified of multilingual experience. • Cost Reduction: Fewer calls routed
known outages, reducing call volumes. to human agents result in up to 25%
Example: Global companies like Vodafone operational cost savings.
Example: Delta Airlines introduced an IVR offer language-specific IVR flows, improving
that automatically provided flight status CSAT across regions. • Customer Effort Score (CES):
updates, cutting down their incoming query Simplifying the interaction flow
volume by 40%. reduces customer effort, directly
boosting CES by 10-15%.
 Traditional vs Conversational IVR
Feature/Aspect Classic IVR Advanced IVR (AI-Powered) Impact on Operations
Natural Language Processing Classic: High abandonment rates due to complex menus.
DTMF-based (touch- (NLP), conversational IVR
Call Navigation
tone, menu-based) Advanced: Lower abandonment, faster query resolution
with conversational input.
Classic: Limited task completion, more queries escalated
Comprehensive self-service (e.g., to agents.
Limited, basic tasks
Self-Service Options payment processing,
(e.g., balance inquiry)
troubleshooting) Advanced: Higher containment, fewer escalations, reduced
operational costs.
Classic: Poor customer satisfaction due to long hold times
and complex menus.
Frustrating, repetitive Seamless, personalized interaction
Customer Experience
menu options based on customer data
Advanced: Improved CSAT with personalized interactions
and fewer menu layers.
Classic: Inefficient routing, high transfer rates, longer AHT.
Basic rule-based routing AI-driven, dynamic routing based
Routing Capability
(e.g., press 1 for billing) on customer intent and history Advanced: Improved FCR with intelligent routing, reducing
agent workload.
Classic: Increased agent workload, manual lookups, errors
Limited, requires in data entry.
Fully integrated, automatic data
Integration with CRM manual data input by
pulling from CRM systems
agents Advanced: Automated, real-time data population for
agents, reducing handling time.
 Traditional vs Conversational IVR
Feature/Aspect Classic IVR Advanced IVR (AI-Powered) Impact on Operations
Classic: Higher after-call work, longer agent post-call time.
Full automation of post-call
Minimal automation
Post-Call Automation workflows (e.g., CRM updates,
(basic ticket generation) Advanced: Reduced after-call time by automating ticketing
follow-ups)
and follow-up tasks.
Classic: High ongoing costs due to labor-heavy operations.
Higher initial setup cost,
Lower initial setup cost,
Cost Efficiency significantly lower operational
higher operational costs Advanced: Lower long-term costs through automation and
costs
reduced agent workload.
Classic: Low IVR containment, requires human intervention
NLP and NLU-based speech for most queries.
None, relies on touch-
Speech Recognition recognition, understanding natural
tone inputs
speech Advanced: Higher containment with real-time, accurate
speech-to-text conversion.

Key Takeaways: Impact of Classic vs Advanced IVR on Operations

Classic IVR Advanced IVR

While cheaper to implement, classic IVR systems lead AI-powered IVR systems enhance efficiency by reducing
to higher operational costs due to more escalations, human intervention, speeding up resolutions with
longer call handling times, and lower containment dynamic routing, and improving self-service options.
rates. The customer experience is often frustrating, This lowers AHT, boosts containment, cuts costs, and
leading to lower satisfaction and higher churn rates. increases customer satisfaction.
 Agent Asist
Overview
Agent Assist uses AI and machine learning tools to support call center agents in real-time by providing contextual suggestions
and information during interactions.
Features

Real-time Recommendations: Knowledge Base Integration: Sentiment Analysis:

Key

Suggests responses or resources based on Pulls relevant data from internal systems to Analyzes customer sentiment to
the conversation context. provide agents with the right answers. guide the agent’s tone and response.
Benefits

Reduced Average Handle Time (AHT): Improved Customer Satisfaction: Agent Training and Support:
Faster resolution with relevant suggestions. Provides seamless and accurate New agents benefit from AI-driven
responses to complex queries. recommendations, reducing ramp-up time.

Banking Call Center: An AI-powered Agent Assist tool pulls relevant loan policy information as soon as the customer mentions a loan
inquiry, ensuring quick and accurate responses.

AHT Reduction: First Call Resolution (FCR): Agent Satisfaction Score:

Key Metrics Decrease in call resolution time. % of issues resolved on the first call Measures how effectively the tool
supports agents.
 Agent Assist: Continuous Learning for Dynamic Agent
Support
Agent Assist tools not only provide real-time support but continuously learn and improve over time, offering dynamic, AI-driven suggestions that
evolve based on new data and feedback from agents.

Real-Time Knowledge Base Updates: Agent Assist tools

Contextual Escalation Recommendations: If the system
continuously learn from ongoing conversations and
detects a certain level of customer frustration, it can
update the knowledge base with new solutions.
recommend escalation to a senior agent.
Advanced Features
Example: If agents frequently struggle with a certain type
Example: Verint’s AI tools suggest real-time escalation when
of inquiry, the AI can prioritize it in future conversations
sentiment analysis identifies high customer frustration.
and offer more accurate, tailored suggestions.

Advanced Use Cases Impact on Call Center KPIs

• Reduced Onboarding Time: Training
time for new agents decreases by 30-
Multi-Channel Support: Agent Assist tools On-the-Job Training: Continuous AI 40%.
integrate with multiple communication learning allows new agents to receive
channels (chat, email, voice), ensuring ongoing training through real-time • Improved First Call Resolution (FCR):
agents have a unified view across channels. suggestions, reducing formal training time. FCR rates increase as agents have
access to real-time, relevant information.
Example: A major retail chain equipped Example: A BPO reduced its training
agents with multi-channel assist tools, duration for new agents by 40% by using AI • Agent Satisfaction: Agents feel less
reducing agent handling time by 30%. assist tools during live calls. overwhelmed by customer queries,
leading to a 15-20% improvement in
agent satisfaction scores.
 Unified Desktop
Overview
A Unified Desktop integrates multiple systems and tools into a single interface for agents, streamlining workflows by minimizing
the need to switch between applications.
Features

Single Pane of Glass Interface: Customer 360 View: Task Automation:

Key

Consolidates CRM, knowledge base, IVR, Provides a comprehensive view of customer Automates repetitive tasks like ticket
ticketing, and communication tools. history, previous interactions, and preferences. creation and follow-ups.
Benefits

Increased Agent Productivity: Consistent Customer Experience: Faster Onboarding:

Reduces time spent switching between Agents have all the information they need in New agents quickly learn to manage tasks
applications. one place to handle queries smoothly. through an integrated interface.

Retail Support Center: A unified desktop integrates chat, email, and voice channels, providing agents with a 360-degree customer view
and automating ticket logging and follow-ups.

Agent Productivity Metrics: Customer Satisfaction (CSAT): System Downtime:

Key Metrics Number of queries handled per hour. Impact of the unified experience on Tracks any system issues or integration
customer feedback. breakdowns.
 Unified Desktop: Data-Driven Insights for Operational
Optimization
Unified Desktop systems not only streamline workflow by consolidating multiple tools but also provide critical data insights that managers can
use to optimize operations.

Real-Time Data Monitoring: Unified Desktop tools provide Predictive Analytics: These systems can forecast issues
real-time analytics on agent performance, customer before they arise by analyzing patterns in customer
interactions, and system health. interactions or system performance.
Advanced Features
Example: Managers can monitor agent availability and Example: If a surge in support requests is detected, the
reassign workloads based on real-time data. system can predict potential bottlenecks and suggest
proactive measures, such as adding more agents.

Advanced Use Cases Impact on Call Center KPIs

• Operational Efficiency: Data-driven
insights help reduce Average Handling
Workforce Management: Unified Desktops Cross-Department Collaboration: Teams Time (AHT) by 15-20%.
allow managers to make informed staffing across departments can share data
decisions based on data from various effortlessly, enhancing customer resolution • First Call Resolution (FCR): FCR
touchpoints (IVR, chat, email). times. improves due to predictive call routing
and better-equipped agents handling
Example: A retail company reduced agent Example: A financial services provider inquiries.
idle time by 20% using predictive staffing reduced its call transfer rate by integrating
recommendations. billing, technical, and CRM systems into a • Customer Satisfaction: CSAT increases
unified desktop. due to faster resolutions and reduced
need for callbacks.
 Speech-to-Text Tools
Overview
Speech-to-text tools convert spoken language into written text, enabling seamless transcription of calls, documentation of
interactions, and support for real-time analysis in call centers. These tools enhance operational efficiency and accuracy in
handling large volumes of verbal communication.

Real-Time Transcription: Speech Recognition Accuracy: Sentiment and Intent Detection:

Features

Converts live conversations into text as they Leverages machine learning and AI to improve the Analyzes tone and keywords within the
Key

occur, supporting immediate accuracy of transcriptions over time, particularly speech to gauge customer sentiment
documentation and processing. in recognizing industry-specific jargon or accents. and intent.

Improved Documentation: Enhanced Call Analysis: Compliance and Security:

Benefits

Automatically generates text records of Enables quicker post-call analysis by making Supports adherence to regulatory
customer interactions, reducing manual conversation data easily accessible for requirements by maintaining accurate and
entry and improving accuracy. review. searchable records of calls.

Financial Services Call Center: A speech-to-text system automatically transcribes interactions between customers and agents, providing
searchable records for quality assurance and compliance checks. This reduces manual logging errors and improved call handling.

Transcription Accuracy Rate: Turnaround Time for Transcriptions: Impact on Compliance:

Key Metrics Measures the percentage of correct words Assesses how quickly the system can Evaluates the system’s role in maintaining
transcribed during calls. convert spoken words into text. accurate records for audits and regulatory needs.
 Speech-to-Text Tools: NLU and NPU (1/3)
Overview
Natural Language Processing (NLP) refers to the ability of machines to read, understand, and derive meaning from human
language. It covers a wide range of tasks, including speech recognition, text analysis, translation, and other language-related
challenges.

Natural Language Understanding (NLU) is a subset of NLP focused on extracting meaning and context from human language,
allowing systems to comprehend, interpret, and act on spoken or written inputs in a way that mimics human understanding.

Intent Recognition: NLU can identify the purpose or intent Contextual Awareness: NLU understands the context of a
behind a customer's query, even when phrased in complex conversation and can carry it over multiple exchanges –
or informal language. responding accurately even when customers switch topics

Example: A customer might ask, “Where’s my order?” or “I Example: If a customer previously asked for information
haven’t received my package yet.” NLU recognizes both as about a product, the system could follow up with
Key Features

an intent to check the order status. questions related to their purchase decision.

Entity Extraction: NLP extracts important data points such Sentiment Analysis: NLP tools analyze the tone and mood
as names, dates, numbers, and keywords from of the conversation to detect customer satisfaction,
conversations. frustration, or urgency.

Example: A system can automatically extract the Example: Detecting frustration in a customer’s voice can
customer’s name and order number from a spoken query. trigger escalation to a live agent.
Speech-to-Text Tools: NLU and NPU (2/3)
Enhanced Self-Service:
Accuracy in Responses:
By leveraging NLU, customers can resolve their issues
NLP ensures that automated systems provide precise,
without agent intervention, leading to higher IVR
relevant answers based on the customer’s query.
containment rates and reduced operational costs.
Benefits

Continuous Learning: Personalization:

NLU and NLP models improve over time by learning These tools help create personalized customer
from each interaction, ensuring responses become experiences by understanding the customer’s history,
more accurate and helpful. preferences, and previous interactions.

Entity Recognition Sentiment Detection Containment Rate

Intent Accuracy
Accuracy Rate Improvement
Measures the Assesses how accurately Measures how many
percentage of correctly the system extracts Monitors how well the queries are resolved
identified intents from critical data points like system can identify within the IVR or chatbot
customer emotions (e.g.,
Key Metrics

customer queries names or order numbers. system without escalating

frustration, satisfaction). to an agent.

Impact: Higher sentiment

Target: Achieving over Impact: Accurate data Impact: A well-
detection rates lead to
90% intent accuracy extraction reduces optimized NLU system
improved customer
leads to higher errors in call handling. can increase
experiences, as systems
containment rates. containment rates by up
can escalate issues when
to 30%.
necessary.
 Speech-to-Text Tools: NLU and NPU (3/3)
Real-World Use Case

Telecommunications Company: A telecom provider implemented NLU-powered IVR to handle routine customer queries such
as checking balances, updating account details, and tracking orders.

Impact: The NLU system resolved 65% of customer queries without human intervention, leading to a 20% reduction in call
volumes to live agents.

Sentiment Analysis: Based on the customer’s tone, the system escalated urgent issues, reducing customer churn by 10%.

How NLU + NLP Transform Call Centers

Better Call Routing Efficient Agent Support 24/7 Customer Support

NLU ensures that customers are NLP tools assist agents by With NLU-driven conversational
FEATURES directed to the appropriate summarizing customer queries IVRs and chatbots, companies can
department based on their specific and offering relevant information in offer round-the-clock service
needs. real-time. without increasing staff.

This reduces call transfers, This reduces agent effort and

Increases customer satisfaction
improves First Call Resolution boosts productivity, allowing
IMPACT (FCR), and cuts Average Handling them to handle more complex
and reduces the need for hiring
additional agents.
Time (AHT). queries effectively.
 Speech-to-Text: Enhancing Compliance and Data
Security
Speech-to-Text tools play a critical role in ensuring regulatory compliance and data security in call centers. Beyond transcription, these tools help
maintain audit trails and ensure that sensitive customer data is protected.

Automatic Redaction: Advanced Speech-to-Text tools Sentiment Analysis and Intent Detection: Tools analyze the
can automatically redact sensitive information, like credit text to identify customer sentiment or potential compliance
card details or Social Security Numbers, in transcripts. risks (e.g., complaints about data breaches).
Advanced Features
Example: A finance company uses speech recognition to Example: Speech-to-text solutions can flag calls for further
redact personal information in real-time to ensure review if they detect negative sentiment related to privacy
compliance with GDPR and HIPAA. concerns.

Advanced Use Cases Impact on Call Center KPIs

• Compliance Adherence: Ensures full
compliance with regulatory frameworks
Regulatory Compliance: Speech-to-text Fraud Detection: Speech-to-text analytics (e.g., GDPR, HIPAA), reducing legal risks.
helps companies adhere to industry tools can detect potential fraud by
regulations by ensuring every interaction is analyzing tone and linguistic patterns • Error Reduction: Automated
accurately recorded and stored securely. during a call. transcription & redaction reduce human
errors, improving data accuracy by 20%.
Example: A healthcare provider uses Example: Financial services companies
automated transcription and redaction use speech analytics to identify fraud risks, • Faster Audits: STS systems facilitate
features to ensure HIPAA compliance while cutting down response time to suspected quick retrieval and review of transcripts
reviewing patient inquiries. fraud by 50%. during internal or external audits,
reducing audit times by up to 25%.
Enhancing Your IVR System with these Technologies

01 02 03
Agent-Enhanced Omnichannel
Seamless Handoff
Workflows Continuity
Agent Assist tools
Unified desktops ensure
Conversational IVRs can integrated into the
that agents have access
transfer calls to agents Unified Desktop enable
to both IVR and other
with Agent Assist, agents to seamlessly
channel interaction
providing all necessary switch between channels
histories for consistent
information upfront. and maintain high-quality
customer experiences.
customer interactions.

Summary of Metrics Across Technologies

Technology Key Metrics

Conversational IVR Containment Rate, Abandonment Rate, CES
Agent Assist AHT Reduction, FCR, Agent Satisfaction Score
Unified Desktop Agent Productivity, CSAT, System Downtime
Identifying Transformation Opportunities with
Technology (1/2)
Each of the tools we’ve discussed not only streamlines operations but also reveals key data points that allow you to identify inefficiencies or areas
ripe for transformation. Your role in daily operations should involve observing patterns and using these tools to improve processes.

Unified Desktop – Improving Workflow Efficiency

Monitor Agent Performance Identify Automation Opportunities Cross-Department Data

Sharing
Look for tasks that can be automated, such as
Regularly review agent data from the desktop
ticket creation or follow-ups. For example, a Unified Desktop allows seamless collaboration
interface to spot bottlenecks. Are agents
sales team found that automating follow-ups between departments (e.g., sales, support),
spending excessive time switching between
through Unified Desktop cut handling times by reducing transfer times and boosting FCR.
tools or handling repetitive tasks?
15%.

Conversational IVR – Enhancing Self-Service Capabilities

Review Call Data Expand Self-Service Options Customer Journey Mapping

Analyze IVR data for low containment or high
Spot repetitive inquiries that could be automated
abandonment rates, which signal inefficiencies. Look at where customers are dropping off in
within the IVR. For example, allowing customers
Simplifying IVR options, like a telecom the IVR system and adjust the flow to simplify
to manage order status or payments through the
company that cut abandonment by 25%, can their experience.
IVR can reduce agent workload by up to 20%.
improve performance.
Identifying Transformation Opportunities with
Technology (2/2)

Speech-to-Text – Analyzing Call Data for Continuous Improvement

Review Call Transcriptions Automation Potential Example:

A retail company used speech analysis to detect
Use Speech-to-Text to analyze customer Identify customer requests that are frequently bottlenecks in their order inquiries, leading to a
feedback. Frequent complaints or long-winded repeated, such as "I forgot my password." These 15% improvement in agent productivity by
conversations are indicators of process can be automated or handled by AI chatbots to updating their FAQs and automated options.
inefficiencies. free up agents.

Agent Assist – Enhancing Agent Effectiveness in Real-Time

Monitor Agent-Tool Interaction Sentiment Analysis Continuous Learning

If the AI detects frequent customer frustration,
Check which suggestions from Agent Assist are
focus on real-time adjustments or training Use insights from the system to identify weak
frequently ignored or lead to poor resolutions.
needs. For example, a BPO reduced escalations spots in agent knowledge and adjust training
This might indicate the tool needs to be
by 20% after using Agent Assist to train agents in programs accordingly.
updated to provide more relevant suggestions.
handling difficult conversations.
 Practical Checklist for Spotting Transformation Opportunities

01 02 03 04
Unified Desktop Conversational IVR Speech-to-Text Agent Assist
Are agents switching between too Do call transcriptions show
Are customers frequently Are agents dismissing AI
many systems? Could automating frequent complaints about similar
bypassing self-service options or recommendations, or is there a
tasks reduce manual entry and issues or lengthy calls due to lack
abandoning calls? pattern of unresolved calls?
handling times? of information?

Action: Propose consolidating Action: Suggest updating FAQs,

Action: Recommend expanding Action: Review the relevance of AI
tasks like ticket management and introducing automated responses
self-service options or updating suggestions, and adjust the
CRM updates into the Unified for frequent queries, or retraining
the IVR menu based on call knowledge base or offer additional
Desktop to improve agent agents based on transcript
patterns. agent training.
productivity. analysis.

Summary of Impact

Unified Desktop Conversational IVR

Improves agent efficiency and reduces handling Increases IVR containment and reduces human
times by identifying process inefficiencies. workload by expanding self-service.

Speech-to-Text Agent Assist

Helps identify repetitive issues and training needs Boosts agent performance and customer satisfaction
through transcription analysis, driving automation. by offering real-time, relevant guidance.
Vendor Spotlight – How Genpact, Teleperformance, and Sutherland
Drive Back-Office Efficiency
Genpact – Leveraging AI for End-to-End Process Automation
Genpact uses proprietary tools such as Genpact Cora to streamline end-to-end business operations, focusing on automating
Overview
repetitive tasks and optimizing decision-making.

Key Features Real-World Example Back-Office Efficiencies

Reduction in Manual Tasks: Automated

Robotic Process Automation (RPA): Use Case: Genpact implemented Cora repetitive workflows such as data entry
Automates back-office processes like to automate the invoicing process for a and reporting.
invoice management, order processing, large retail client, reducing manual data
and data entry. entry errors by 30% and cutting Improved Decision Making: AI-driven
processing time by 25%. insights allow businesses to predict
AI and Machine Learning: Enables demand, optimize resource allocation,
predictive analytics for tasks such as Impact: Improved invoice accuracy, and streamline financial operations.
supply chain management and financial reduced processing time from days to
forecasting. hours, and enhanced cash flow Cost Savings: A 20-30% reduction in
management for the client. operational costs through automation
and optimized resource use.
Vendor Spotlight – How Genpact, Teleperformance, and Sutherland
Drive Back-Office Efficiency
Teleperformance – Enhancing Customer Support with AI-Driven Analytics
Teleperformance utilizes its proprietary platform TP Observer and other AI-powered tools to optimize both front-end customer
Overview service and back-office operations like quality assurance and compliance.

Key Features Real-World Example Back-Office Efficiencies

Automation in Quality Assurance:

Automated review of call transcripts and
Speech and Text Analytics: Analyzes Use Case: Teleperformance used TP customer interactions to ensure
interactions in real time to detect Observer for a telecom client to compliance and performance metrics
sentiment, compliance issues, and automate the compliance review are met.
overall agent performance. process, cutting review times by 40%
while increasing compliance adherence Real-Time Data Analysis: Continuous
Automated Quality Assurance: by 20%. analysis of agent-customer interactions
Automates the QA process by analyzing to optimize workflow and provide real-
call transcripts and flagging non- Impact: Reduced the need for manual time feedback to agents.
compliant conversations. reviews, improved agent performance
through real-time feedback, and Improved Compliance and QA:
enhanced overall compliance with Reduced manual labor associated with
regulatory requirements. quality checks and compliance audits,
boosting operational efficiency.
Vendor Spotlight – How Genpact, Teleperformance, and Sutherland
Drive Back-Office Efficiency
Sutherland – Process Mining and Intelligent Automation for Workflow Optimization
Sutherland is a leader in applying process mining and intelligent automation to optimize back-office workflows in industries such
Overview as finance, healthcare, and retail.

Key Features Real-World Example Back-Office Efficiencies

Process Optimization: Process mining

helps identify bottlenecks and
Process Mining: Analyzes business Use Case: Sutherland applied Robility inefficiencies, allowing companies to
processes to detect inefficiencies and to automate the claims processing for a streamline workflows.
bottlenecks in workflows. healthcare provider, reducing the
processing time by 50% and improving Time Savings: Automating processes
Intelligent Automation: Automates claim accuracy by 35%. like claims handling and invoice
repetitive and manual tasks, from generation drastically cuts down
claims processing to order Impact: Enabled faster turnaround processing time.
management. times for claims, reduced errors due to
automation, and enhanced client Error Reduction: By automating data
satisfaction by speeding up entry and other manual tasks,
reimbursement processes. Sutherland achieved a 35% reduction in
errors for their healthcare client.
 IVR Containment – Reducing Human Intervention and Improving Call
Center Performance
What is IVR Containment?
IVR containment refers to the percentage of calls that are handled entirely within the IVR system, without the need for escalation to a live agent.

Objective The goal is to resolve customer queries efficiently through self-service, reducing the volume of calls that require agent assistance.

Self-Service Menus: Natural Language Processing

Contextual Routing:
Offering customers intuitive, (NLP):
Using AI to anticipate customer
Key Components of well-designed self-service Conversational IVR allows
needs based on prior
IVR Containment options to handle frequent customers to state their issue in
interactions and offer relevant
queries (e.g., account balance, their own words, enhancing IVR
solutions.
password reset). containment.

Decreasing Call Volume Impact on Agents Common Self-Service Scenarios

How IVR Containment
As more queries are resolved Agents can focus on more complex Automate order tracking, billing
Reduces Human
through the IVR, the volume of and high-value queries, increasing inquiries, and service updates to
Intervention calls reaching agents is their effectiveness and reducing minimize customer calls and agent
significantly reduced. burnout. dependency.
 Data-Driven Insights on IVR Containment’s Impact on Performance

Companies with higher containment

rates see a 25-30% reduction in agent-
handled calls.

Example: A telecom provider's Containment Rate

conversational IVR boosted Higher containment leads to cost
containment by 30% and cut agent call savings as fewer agents are needed to
volume by 20%. handle routine calls.
Cost Per Call Reduction
Data Point: Studies show that every
10% increase in containment can
IVR containment improves FCR as First Call Resolution reduce operational costs by 10-15%.
common queries are handled without
transfer delays.
(FCR) Improvement
Faster, automated solutions reduce
Data Point: A well-optimized IVR can Customer Satisfaction frustration and wait times, improving
increase FCR by up to 20%, as
customers get immediate solutions.
(CSAT) Improvement CSAT.

Data Point: Companies that improve

IVR containment see a 10-15% boost in
CSAT, as customers experience
quicker resolutions.
 Data-Driven Results – Real-World Examples of Improved IVR
Containment
Success Story 1: Verizon – Enhancing Containment with AI and NLP Success Story 2: Delta Airlines – Optimizing IVR for Travel Inquiries

Overview: Delta revamped their IVR system to manage common flight

Overview: Verizon implemented a conversational IVR powered by NLP inquiries such as flight status, cancellations, and upgrades.
to handle common customer queries (like billing and service outages).
Results:
Results: • Containment Rate: Increased from 40% to 60%, drastically
• Containment Rate: Increased from 50% to 65% in six months. reducing agent involvement in routine inquiries.
• Call Volume Reduction: 25% reduction in calls routed to agents. • FCR Improvement: Increased FCR by 15% due to fewer transfers
• Cost Savings: Saved $2 million annually in operational costs due to and faster call resolutions.
fewer calls requiring live agents. • Operational Efficiency: Reduced average handling time by 10% and
• CSAT Increase: 12% improvement in CSAT as customers reported saved over $1.5 million annually in labor costs.
faster resolution times through self-service. • Customer Feedback: Customers reported less frustration with wait
times, leading to a 10% boost in CES (Customer Effort Score).

Key Takeaways for Call Center Efficiency

• Higher IVR Containment Rates = Fewer Agent Interactions: When IVR is optimized, routine calls are handled entirely through self-service,
freeing agents to focus on complex issues.

• Cost Savings: Every percentage increase in containment results in significant cost reductions, often in the range of 10-20%.

• Performance Improvements: By optimizing containment, companies see improvements in AHT, FCR, and CSAT, driving both efficiency and
customer satisfaction.
Understanding How Unified Desktop Tools Ensure Smooth
Workflows (1/2)
Inputting Structured Data
Unified Desktop tools rely on structured, consistent data inputs to function effectively. When agents and automated systems provide clear,
accurate information, it ensures smoother workflows and better customer outcomes.

Specific Examples of Inputs that Ensure Success

Binary Selections for Automated Comments Predefined Automated Responses

Overview: Using binary fields like yes/no or resolved/unresolved Overview: Unified Desktops can provide agents with predefined,
allows agents to quickly select the status of a task or query. automated responses for common customer queries.

Example: When closing a ticket, agents can select “Resolved” or Example: When customers request order tracking information,
“Escalated.” This data helps determine how many cases are the system can automatically pull the order details and provide
handled successfully on the first call (FCR metric). the agent with an automated response.

Impact: Simplifies data entry and ensures uniform reporting, Impact: Reduces manual workload for agents, speeds up
making it easier to track agent performance and identify response time, and ensures accuracy in communication,
bottlenecks. improving CSAT (Customer Satisfaction) scores.
 Understanding How Unified Desktop Tools Ensure Smooth
Workflows (2/2)
Dropdown Menus for Common Customer Queries Real-Time Data Sync for Omnichannel Interactions

Overview: Instead of manually typing out common issues,

Overview: Unified Desktops automatically pull data from all
dropdown options standardize the process.
customer touchpoints—chat, email, phone—into one central
hub.
Example: A dropdown menu with pre-populated queries such as
“Billing Issue,” “Technical Support,” or “Order Status” helps
Example: If a customer has been interacting with a chatbot
agents categorize calls efficiently.
before speaking with an agent, the chatbot conversation is
automatically pulled into the agent’s dashboard.
Impact: Reduces time spent on data entry, ensuring consistency
in categorizing and analyzing customer issues. This helps in
Impact: Eliminates the need for customers to repeat themselves,
generating insights into common queries and optimizing IVR
improving customer experience and agent efficiency.
menus for future interactions.

Impact of Well-Populated Unified Desktop Tools on Performance

Consistency Operational Efficiency Data-Driven Insights

When all agents use the same data fields Automating routine tasks and standardizing With structured data from binary selections,
(binary selections, dropdown menus), the inputs reduces AHT by up to 20% and dropdown menus, and automated responses,
data collected is consistent, making reporting increases FCR by 10%, as agents focus more managers can track common issues, agent
and analysis more accurate. on resolving complex issues rather than performance, and workflow bottlenecks,
manual tasks. helping drive continuous improvement in
processes.
 Summary: Critical Milestones in the Customer Call Flow

01 02 03 Call Routing and Queue 04 Initial Customer-Agent

Pre-Call Preparation – IVR Interaction –
Management – Interaction –
Customer Data Identification Initial Contact and Routing
Getting the Right Agent Building Rapport

Milestone: Milestone:
Milestone: Milestone:
The first point of contact for Once the customer’s query is
Before the call even begins, Once connected, the agent
many customers is the IVR understood, they must be routed
customer data is critical for greets the customer and verifies
system, where they select to the appropriate agent or
effective call routing. their information.
options related to their query. department.

08 Analytics and Continuous 07 06 05 During the Call –

Post-Call Wrap-Up – Call Resolution –
Improvement – Real-Time Speech-to-Text
Data Entry and Follow-Ups Automating Routine Tasks
Call Monitoring and Feedback Transcription

Milestone: Milestone:
Milestone: Milestone:
After all customer interactions, After the call ends, agents must
After understanding the As the conversation progresses,
it's crucial to analyze update systems with the call
customer’s problem, the agent capturing information accurately
performance and identify areas outcome, schedule follow-ups,
takes action to resolve it. is essential.
for future improvement. or send notifications.
 Critical Milestones in the Customer Call Flow – Opportunities for
Transformation (1/2)

01 02 03 Call Routing and Queue 04 Initial Customer-Agent

Pre-Call Preparation – IVR Interaction –
Management – Interaction –
Customer Data Identification Initial Contact and Routing
Getting the Right Agent Building Rapport

Opportunity for Transformation:

Opportunity for Transformation:
Opportunity for Transformation: Opportunity for Transformation:
Advanced Conversational IVR:
Implementing AI-driven conversational Automation with Pre-Populated Data:
Automation Solution: Integrating CRM Intelligent Call Routing with AI: AI-
IVR can transform the customer Automated systems can pre-populate
systems with IVR and automated powered call routing can dynamically
experience by understanding natural customer data for the agent, reducing
customer identification can speed up assign customers to the most qualified
language. the need for manual information
call handling. agent based on real-time data (such as
gathering.
current workload, customer history, or
How: Instead of touch-tone systems,
How: Automation solutions pull sentiment).
conversational IVR enables customers How: The system pulls customer
customer data (e.g., phone number,
to state their problem in their own details (from previous interactions,
past interactions) and instantly How: AI evaluates the complexity of the
words, improving self-service success product usage, etc.) and presents them
populate the agent’s dashboard with call and routes high-priority calls to
rates. to the agent in real-time.
relevant information. senior agents or technical experts.
Impact: Natural language Impact: This can reduce call handling
Impact: This eliminates the need for Impact: This reduces transfer rates and
understanding reduces call transfers by time by 10% and allow agents to spend
manual lookups and reduces Average improves First Call Resolution (FCR) by
20%, increases IVR containment, and more time focusing on resolving the
Handling Time (AHT) by 10-15%. 15-20%.
provides smoother transitions to agents issue instead of data collection.
when necessary.
 Critical Milestones in the Customer Call Flow – Opportunities for
Transformation (2/2)

05 During the Call – 06 07 08 Analytics and Continuous

Call Resolution – Post-Call Wrap-Up –
Real-Time Speech-to-Text Improvement –
Automating Routine Tasks Data Entry and Follow-Ups
Transcription Call Monitoring and Feedback

Opportunity for Transformation:

Opportunity for Transformation: Opportunity for Transformation:
Opportunity for Transformation:
AI-Powered Call Analytics: Analyzing
Speech-to-Text for Real-Time Automation of Routine Tasks:
Post-Call Automation: Automated calls through AI-driven analytics tools
Transcription: Using speech-to-text Automating repetitive tasks like
post-call wrap-up processes can can help management understand
technology, the call is transcribed in resetting passwords, generating
handle ticket generation, follow-up trends, spot recurring issues, and
real-time, helping agents focus on the invoices, or scheduling follow-ups can
scheduling, and data entry. identify training needs.
conversation while ensuring all details streamline call resolution.
are recorded.
How: After the call, an automated How: Speech analytics combined with
How: With an integrated automation
system logs the call, updates CRM customer sentiment analysis can
How: Real-time transcription solution, agents can trigger automated
records, schedules follow-ups, and highlight areas where the customer
automates the note-taking process, workflows (such as sending an email
sends any necessary confirmations to journey could be improved or where
reducing the risk of human error and confirmation or updating CRM records)
the customer. additional agent training may be
ensuring critical information is captured with a single click.
necessary.
for future reference.
Impact: This can reduce after-call work
Impact: This reduces the need for
by up to 40%, allowing agents to move Impact: Continuous analysis and
Impact: This improves accuracy and manual follow-up work, improving
quickly to the next customer feedback can drive ongoing
reduces post-call work, cutting agent efficiency and allowing agents to handle
interaction. improvements in FCR, AHT, and
after-call time by 30%. more calls.
customer satisfaction scores.
Importance of Vendor-Specific Tools in BPO Operations
Why Vendor-Specific Tools Matter
In the highly competitive BPO industry, vendor-specific tools are key to driving operational efficiency, reducing costs, and delivering a superior
customer experience. These tools are tailored to the specific needs of BPOs, enabling them to optimize workflows and handle large volumes of
customer interactions efficiently.

Custom-Fit Solutions
Tools designed by vendors like Genpact, Teleperformance, and Sutherland are built to meet the exact needs of their operations, enabling:

Scalability Automation Data-Driven Insights

Handling high call volumes Reducing reliance on manual

Enabling real-time analytics
and adapting to seasonal processes, which cuts costs
that drive decision-making.
demand fluctuations. and improves speed.

End-to-End Process Optimization: Tailored to Business Goals:

How Vendor-Specific By integrating speech-to-text, process mining, and These tools are configured based on each BPO’s unique
Tools Drive Efficiency automation solutions, BPOs can automate repetitive needs, whether it's improving compliance, reducing
tasks, enhance decision-making, and offer better operational costs, or boosting First Call Resolution
customer experiences. (FCR).
Real-World Applications by Genpact, Teleperformance, and
Sutherland (1/3)
Genpact – Leveraging Process Mining and Automation for Workflow Efficiency

Tool Genpact Cora integrates process mining with automation to streamline end-to-end operations.

Process Mining Automation

How It Works Analyzes workflows to identify inefficiencies and bottlenecks. Once inefficiencies are identified,
For instance, Genpact uses process mining to understand automation takes over routine tasks such as
how call escalations and resolutions are managed, updating CRM systems or handling
pinpointing areas for improvement. invoicing.

Real-World
For a financial services client, Genpact Cora This combination of process mining
Example and automation allowed the client to
automated the entire invoicing process, cutting IMPACT reduce operational costs by 20%,
down manual processing by 30% and reducing while improving accuracy and
errors by 25%. turnaround times
Technology
Real-World Applications by Genpact, Teleperformance, and
Sutherland (2/3)
Teleperformance – Enhancing Quality Assurance with Speech-to-Text and Automation

TP Observer integrates speech-to-text capabilities with automated quality assurance tools to

Tool
ensure regulatory compliance and improve agent performance.

Speech-to-Text Automation in Quality Assurance

How It Works Automatically transcribes customer-agent conversations in Automatically flags non-compliant
real-time, enabling faster reviews and automated sentiment conversations or underperforming agents,
analysis. reducing the need for manual reviews.

Teleperformance implemented TP Observer to

Real-World handle the compliance review process for a
Example telecom client. By automating the transcription
This resulted in a 15% improvement
and compliance checks, Teleperformance IMPACT in agent performance and a 10%
boost in First Call Resolution (FCR).
reduced review time by 40% and improved
compliance adherence by 25%.
Technology
Real-World Applications by Genpact, Teleperformance, and
Sutherland (3/3)
Sutherland – Optimizing Customer Interactions with Conversational IVR and Automation

Sutherland Robility combines conversational IVR and automation to manage high call volumes
Tool
while reducing the need for human intervention.

Conversational IVR Automation

How It Works
Sutherland uses AI-driven conversational IVR to Automated workflows handle post-call tasks such
handle routine inquiries (e.g., order status, as updating customer records and triggering
payment issues) without escalating to human follow-ups, allowing agents to focus on more
agents. complex issues.

Real-World A retail client of Sutherland saw a 35% This translated to a 20% reduction
Example increase in IVR containment after deploying in operational costs and improved
their conversational IVR system, reducing the IMPACT agent productivity by allowing them
Technology volume of calls that required human agents. to focus on high-priority issues.
 Vendor-Specific Tools: Key Benefits Across BPO Operations
Speech-to-Text and Quality Conversational IVR and
Process Mining and Automation
Assurance Automation
Example

Genpact’s tools analyze and automate Teleperformance uses speech-to-text to Sutherland’s conversational IVR handles
workflows to reduce errors and increase streamline compliance reviews and routine inquiries, reducing human
operational efficiency. automate agent performance monitoring. intervention.
Impact

Improved process accuracy and a Reduced manual review time by

Improved IVR containment by 35%,
reduction in manual workload by up 40% and increased compliance
lowering operational costs by 20%.
to 30%. adherence by 25%.

Summary of Metrics Across Technologies

Speech-to-Text, Process Mining, and Automation

Vendor-Specific Tools Are Game-Changers:
Lead the Charge:
Genpact, Teleperformance, and Sutherland
These tools streamline workflows, reduce costs, and
demonstrate that tailored tools allow BPOs to
improve customer satisfaction through faster
automate, optimize, and scale operations efficiently.
resolutions and improved compliance.
Poll

What is the primary objective of

your current IVR System?

Customer Cost Call Revenue

Satisfaction Reduction Deflection Generation
Key Metrics to Measure IVR Success

Measure what matters: Optimize call containment, FCR, and customer satisfaction for maximum impact.

1 2 3 4

Customer
Call Containment Average Call First Call Satisfaction
Rate Duration Resolution (CSAT)

The percentage of calls Use post-call

that are resolved without Shorter IVR interactions A higher FCR rate surveys to
needing human indicate efficient system indicates that the gauge
intervention design IVR is resolving satisfaction
issues effectively. after IVR
interactions.

46
Key Examples

Bank of America Verizon Wireless Zappos

Bank of America, was facing a challenge with a high One of the largest telecommunications companies Zappos, a well-known online retailer, recognized
volume of calls being handled by live agents. in the U.S., was facing a significant issue with its for its customer-centric approach, noticed that
Despite having an IVR in place, many customers IVR system. Many customers were hanging up their CSAT scores from IVR interactions were not
were still bypassing self-service options and opting before completing their interactions, leading to high as high as expected, compared to interactions
to speak with an agent. call abandonment rates with live agents.

Success Story: They expanded self service Success Story: Simplifying the call flow by Success Story: Introduced IVR post-call surveys,
options, simplified the navigation process, reducing no. of menu options, moving high-priority simplified the menu, introduced more
Personalization and enhanced NLP and frequently used options to the top, personalization within IVR system and continuous
implementing NLP and shortening length of monitoring for customer satisfaction.
prompts
Poll

Which metric do you currently

track to assess IVR performance?
Call
First Call Average Call Customer
Abandonment
Resolution Duration Satisfaction
Rate
Data-Driven Improvements for IVR
Systems
Leverage data insights to transform your IVR system from a cost center into a strategic asset

Analysing Call Drop off Improving Call Routing

Rates
Efficiency
Identifying points in the IVR journey where
customers abandon calls. Reducing transfer times and improving
routing accuracy through data.
Failure Example: A retail company had high
Example: Verizon reduced its call
call drop-off rates during promotional
routing errors by 40% after analyzing data
periods. They analysed the data and identified to improve its IVR's natural language
that the promotional menu options were processing capabilities
unclear. Simplifying the options reduced
drop-offs by 30%.

Personalization Through
Leveraging Customer
Data
Feedback
Using IVR surveys and feedback forms to Personalizing the IVR experience by using
drive system improvements. customer history.

Success Story: Delta Airlines used IVR Success Story: Netflix incorporated
customer feedback to simplify flight customer data into its IVR system to offer
status checks, leading to a 25% more personalized assistance, which
improvement in IVR usage increased customer retention by 18%.

49
Poll

Which of the following areas do you think

holds the most potential for improving your IVR
system?

Call Routing Call Customer

Personalization
Efficiency Containment Feedback
Building a Roadmap for
Continuous IVR Improvement
Continuous Monitoring & Implement Small, Employee Training
Feedback Loops Iterative Changes and Adaptation
Real-Time Monitoring: Use analytics A/B Testing: Test small tweaks to see Update Agents: Ensure agents are
to review system performance daily or the impact on KPIs. trained on new IVR changes and
weekly. Prioritize Quick Wins: Focus on understand when customers need to
Customer Feedback: Collect feedback changes that yield the most significant transition to a live representative.
through post-call surveys or sentiment impact in the shortest time. Cross-Functional Workshops:
analysis to identify gaps. Automate Updates: Set up systems to Regular workshops for employees
Agent Input: Regularly gather insights automate frequent updates or adapt across departments to stay aligned
from agents who handle IVR calls. call flows based on live feedback with changes and IVR enhancements.

01 02 03 04 05 06 07
Define Success Conduct Regular Leverage AI & Machine Plan for Scalability
Metrics Audits Learning
Set Key Performance Indicators Call Flow Evaluation: Regularly Personalization: Use AI to personalize Assess Future Needs: Plan IVR
(KPIs): Define what success looks assess IVR call flows and scripts call flows based on caller history or improvements with growth in mind—
like for your IVR system (e.g., First for relevance and ease of use. preferences. consider additional languages, new
Call Resolution (FCR), Average Cross-Functional Audits: Involve Predictive Routing: Leverage machine service lines, or expanded capacity.
Handle Time (AHT), Abandonment teams from IT, operations, and learning to predict the best outcomes Cloud Integration: Integrate cloud-
Rate). customer service for based on real-time caller data. based solutions for flexibility and
Track Call Containment: Monitor comprehensive reviews Natural Language Processing (NLP): scalability as call volumes fluctuate
how effectively the IVR is resolving Enhance the IVR with speech
customer issues without needing recognition for better interaction and
an agent fewer frustrations.
Poll

Which of the following is the most critical

factor for ensuring continuous IVR
improvement in your organization?
Implementing Scalability
Regular Audits
Setting and AI & Machine Planning and
and Feedback
Tracking KPIs Learning for Cloud
Collection
Personalization Integration
Success Story: British Airways
British Airways is the UK's largest international airline, known for its extensive global network and premium travel
experience. It serves millions of passengers annually, connecting major cities across the world.

Long Call Inconsistent Declining Customer

Challenge High Call Volumes
Resolution Times Customer Experience Satisfaction

Phase 1 : Planning Phase 2: Implementation Phase 3: Adoption

Customer Data Analysis: British Airways Predictive Analytics Integration: They Employee Training: Customer service
analyzed frequent flyer behavior and call integrated predictive analytics into their IVR agents were trained on the updated system
patterns to identify key pain points and system, allowing it to anticipate frequent flyer to ensure seamless transitions when
common inquiries. needs and offer personalized options based human intervention was necessary.
Customer Communication: They informed
Approach Goal Setting: The airline set clear objectives
to reduce call times, enhance
on travel history.
AI-Driven Call Flows: Machine learning
frequent flyers about the enhanced system,
highlighting the improved experience with
personalization, and improve customer algorithms were used to refine call flows, faster, more personalized interactions.
satisfaction through predictive analytics and routing calls more efficiently and tailoring Monitoring and Feedback: The airline set
AI-driven solutions. responses based on individual caller profiles. up continuous monitoring of IVR
Technology Selection: They evaluated Continuous Testing and Optimization: They performance, using customer feedback to
potential technologies, focusing on predictive conducted A/B testing and real-time make iterative improvements, ensuring
analytics and machine learning for smarter adjustments to optimize call flows and ongoing system optimization and customer
call routing and personalized experiences. response times satisfaction.

1st Call Decrease in

Reduction Improvement
Impact 30% in Call Time 20% in CSAT 15% Resoluti- 25% abandonem-
on ent rates
The IVR Improvement
Challenge
The IVR Improvement
Challenge
Role Playing & Simulation

Divide into Teams: Each team represents

one department :
• Customer Service
Scenario • IT Roles within
Setup
• Operations
• Marketing each group
• Sales
Each department has received different
customer feedback and data from the IVR
system.

Each team represents a different department of a • Customer (Personal Experience Role): Each
growing company that recently launched a new team member will take turns sharing a real-
IVR system to handle customer service queries. life IVR experience from their personal life.
The system has been running for 6 months, and The group will evaluate this experience as if it
each team is tasked with assessing its success were feedback from a customer.
and proposing improvements.
The twist: Each participant will imagine Groups • Data Analyst: One person in each group will
analyze the metrics and performance of the
themselves as a "customer" who has had a
current IVR system based on fictional data
personal experience with the company's IVR
system while trying to solve a frustrating provided.
problem in their daily life (e.g., billing issue, • System Designer: One person will represent
service request, or travel booking). This the IVR design team and explain how the
personal experience will form the basis of their system was built, detailing its logic and flow.
analysis.
 Key Discussion Points for Each Group

Task Breakdown:
Step 1: Share Personal
Experiences (10 minutes): Each
participant shares an example of
their frustration or satisfaction with
an IVR system from their own life,
Scenario Challenge: describing the pain points and
The customer in each team has successes.
reported difficulties navigating the Step 2: Analyze IVR Performance
IVR system (e.g., lengthy wait times, (20 minutes): Using fictional
unhelpful automated responses, or metrics, each team will review the Debrief & Discussion:
getting disconnected). The team performance of their department's Final Presentation (5 minutes per
Key questions:
must assess: IVR system, analyzing where it team):
• How successful was the IVR succeeds and fails in meeting A brief summary of their
• How did personal experiences
system in resolving their issue? customer needs. customer’s IVR experience.
Step 3: Propose Data-Driven shape the analysis?
• Which key metrics (e.g., call Key performance metrics
Improvements (15 minutes): Based • Were there similarities in the
containment rate, customer analyzed.
on the metrics and personal types of frustrations faced
satisfaction, average handling Proposed improvements and how
feedback, the team will propose 2-3 across different teams?
time) indicate its success or they will be measured for success.
specific, measurable improvements • How did the proposed
failure? improvements align with data-
for the IVR system, such as updating
• What data-driven improvements the flow, improving voice driven decision-making?
can be made to enhance the recognition, or offering callback
system's performance? options.
Thank You
Module 2

Building deeper understanding of

core modelling techniques used in
AI/ML and identifying key drivers
that are used in these model
Session 2
Predictive Model Workflow

Cleaned Data Predictive

Data Data Feature Generate
Model Train Model
Collection Preprocessing Extraction Predictions
Development
Transformed
Data

1. Predictive models help

forecast key metrics, like
call volume and customer
churn.
2. Enables better staffing,
resource allocation, and
1. Call centers generate large handling peak times.
amounts of data from 3. Can optimize customer
interactions. routing based on predicted
2. Data sources include IVR sentiment or urgency
logs, call recordings, agent 1. Call center data can be
performance, and messy, with missing values,
customer feedback. outliers, and inconsistent 1. Not all data points are
3. Ensures data integrity and formats. useful—focus on features
completeness for accurate 2. Preprocessing ensures that that impact operations.
predictions. data is clean and ready for 2. Key features could include
model development. call duration, customer
3. Automating preprocessing sentiment, agent
(e.g., data cleansing and performance, or resolution
normalization) can time.
streamline operations. 3. Helps identify what drives
customer satisfaction and
operational efficiency.
Data Collection & Processing
Data Storage & Predictive Model
Data Gathering
Transformation Development

Customer Interaction Data Step 1: Model Algorithm

1. Algorithm Selection:
Call Recordings IVR Data • Linear Regression
Cloud Based Storage • Decision Trees
• Random Forests
Chat Transcripts Emails & Tickets • Support Vector Machines (SVM)
• Neural Networks
2. Supervised vs Unsupervised Learning
CRM Data
Step 2: Model Training

CRM Systems 1. Training Process:

• Model learns from historical data.
Operational Data • Adjustment of internal parameters
(weights
Agent Performance 2. Loss Function:
Call Logs • A mathematical function used to
Metrics
measure the difference between
the model's predictions and the
Queue Management Data actual outcomes
Data Warehouse

Step 3: Model Validation

Step 4: Model Deployment

Customer Feedback
Step 5: Monitoring & Retraining
Customer Sentiment On-Premise Servers
Surveys
Analysis
 Predictive Modeling Impact
• Predict customer intent, satisfaction, and churn based on voice
Genesys
1 IVR Systems patterns. Nuance
• Provide insights into customer interactions in real-time. Cloud

• Predict call volumes to optimize staffing levels during peak time

2 WFM Tools • Reduce customer wait times by ensuring adequate staffing. Verint NICE IEX

• Predict customer behavior such as churn and upsell opportunities. Salesforce

3 CRM Systems • Enable proactive customer service based on historical interactions. HubSpot
Einstein

• Predict customer intent for more accurate responses.

Google
4 Chatbots & VA • Escalate cases to human agents when necessary, improving resolution IBM Watson
times. Dialogflow

5 Customer Sentiment
• Analyze tone and sentiment in real-time to adjust agent responses. CallMiner Clarabridge
• Monitor trends in customer emotions across call centers. Eureka

• Automatically route calls based on customer history and agent

Zendesk
6 Automated Call Routing expertise. Five9
• Improve first-call resolution rates by routing to the most suitable agent. Talk

• Predict customer intent, satisfaction, and churn based on voice

NICE Verint
7 Speech Analytics patterns.
• Provide insights into customer interactions in real-time. Nexidia Speech

• Flag suspicious customer behavior during calls in real-time. NICE

8 Fraud Detection • Detect and prevent fraud using voice biometrics and behavioral Pindrop
Actimize
patterns
Success Story - Vodafone
American Express (AmEx), one of the world’s leading financial services corporations, manages a high volume of customer interactions across its call
centers globally. AmEx faced challenges in delivering consistent customer experiences, reducing Average Handling Time (AHT), and improving First
Call Resolution (FCR) rates. To stay competitive, AmEx implemented AI/ML technologies to transform its call center operations and deliver superior
customer service.

Inconsistent High Average Low First Call Agent Resistance to

Challenge Customer Experience Handling Time (AHT) Resolution (FCR) AI Adoption

Phase 1: Assessment & Planning Phase 2: AI/ML Implementation Phase 3: Training & Optimization

• Conducted a detailed audit of • Deployed AI-driven predictive call routing to • Retrained agents to collaborate
existing customer service match customers with the best agents based on with AI tools, emphasizing that AI
processes. real-time data would assist, not replace, them.

Approach • Identified inefficiencies in call

routing and agent performance
• Integrated Natural Language Processing (NLP) to
analyze customer sentiment and guide agent
• Offered continuous support for
agents on using AI effectively to
management. responses in real-time improve customer interactions.
• Established KPIs to track success: • Used machine learning models to enhance • Deployed AI-powered quality
AHT, FCR, CSAT (Customer personalization of customer service, predicting assurance tools to automatically
Satisfaction). needs based on historical data evaluate calls.

Operational
Increase in Reduction Increase in
Impact $10M Cost 25% FCR 20% in AHT 17% CSAT
Savings
Core Modelling
Techniques
 Core Modeling Techniques
Core modeling techniques in AI/ML encompass the algorithms and methods used to analyze data, build models, and make predictions.
Overview These techniques form the foundation of most machine learning workflows across various industries, including call centers. By employing these
models, call centers can make data-driven decisions, optimize resources, and enhance customer interactions

Supervised Unsupervised Reinforcement Learning

• Supervised learning is a type of machine • The model is trained on data that has no
Types of Modelling learning where the model is trained on labeled outcomes
Techniques labeled data, meaning that each training • The model must discover patterns, AI and machine learning algorithms are
example has both input data (features) and groupings, or structure within the data on its used to classify data dynamically based on
the correct output (target). own patterns in the data
• The model learns to map inputs to the • There is no explicit "correct" answer
correct outputs & is later tested on new data provided during the training phase

1. The model is given a dataset with input- 1. The model is given a dataset with inputs 1. There are no fixed input-output pairs, but
output pairs. only, without corresponding outputs or instead, the model takes actions in the
2. It learns the relationship between the input labels environment and receives feedback
Process features and the target variable. 2. It learns the inherent structure within the 2. It learns the relationship between its actions
3. Once trained, the model can predict the input data without the need for target and the resulting rewards or penalties
target for new, unseen inputs variables 3. Once trained, the model can make decisions
3. Once trained, the model can group or in new, unseen environments to maximize
classify new, unseen inputs into categories cumulative rewards over time

• Linear Regression • Clustering • Q-Learning

• Logistic Regression • Principal Component Analysis (PCA) • Deep Q-Networks (DQN)
Common • Decision Trees
Techniques • Random Forests
• Support Vector Machines (SVM)
• K-Nearest Neighbours
 Supervised Learning
In supervised learning, algorithms are provided with a set of input-output pairs (labeled data), and the model’s objective is to
Overview learn a mapping function that connects inputs to the correct outputs.

Customer Churn Prediction Agent Performance Evaluation

1. Description: Predict whether a customer is likely to stop using a service 1. Description: Classifying agents based on performance metrics (like
based on their interaction history, demographics, & past behavior average handling time, customer satisfaction, or resolution rate) to
2. How It Works: The model is trained on historical data where the identify top-performing and underperforming agents.
outcome (churn or no churn) is already known. Based on features like 2. How It Works: A model can be trained on past performance data of
How it Works customer sentiment, interaction frequency, and service issues, it agents where the output (performance rating or classification) is
predicts if a customer will leave. known. It predicts future performance, enabling personalized training
3. Use Case: Call centers can use this to focus retention efforts on high- programs for agents.
risk customers by offering proactive support or special incentives. 3. Use Case: Managers can allocate tasks better and provide targeted
training or bonuses based on predicted agent performance.

Linear Regression Decision Trees

A regression algorithm that models the A tree-like model that makes decisions
Key Algorithms relationship between a dependent based on input features, navigating
variable and one or more independent through decision nodes until reaching
variables. This technique is crucial for an output.
predicting continuous outcomes.

Predicting call duration based on factors like Identifying whether a call is likely to be escalated
customer type, issue complexity, and agent based on customer tone, issue type, and call
experience. history.
 Unsupervised Learning
Unsupervised learning utilizes data without labeled outputs. The model seeks to identify hidden patterns or intrinsic
Overview structures in the data without explicit instructions on what to look for

Customer Segmentation Agent Performance Clustering

1. Description: Grouping customers based on behavior, preferences, or
interaction patterns to better understand and serve different customer 1. Description: Grouping agents based on their performance metrics,
groups. such as average call handling time, customer satisfaction scores, or
2. How It Works: Algorithms like K-Means Clustering or Hierarchical resolution rates.
Clustering can automatically identify natural groupings in customer 2. How It Works: Unsupervised learning can identify clusters of agents
How it Works with similar performance patterns, which may help managers provide
data based on factors such as call frequency, purchase history, or
interaction satisfaction. targeted training or identify top performers.
3. Use Case: Call centers can use these segments to create personalized 3. Use Case: Call center managers can group high-performing agents for
customer experiences, tailor marketing efforts, or offer different levels critical calls, while underperforming agents can receive specialized
of service based on customer behavior (e.g., VIP customers vs. training to improve their skills.
infrequent users).

Principal Component Analysis K-Means Clustering

Key Algorithms A popular clustering algorithm that

A dimensionality reduction technique that
divides the data into 'K' groups based on
identifies the most significant features in
similarity
large datasets

Simplifying complex customer interaction data to Grouping customers based on similar complaints
determine which factors most impact CSAT or FCR or interaction histories to tailor service offerings
 Reinforcement Learning
Reinforcement learning involves training an agent to make decisions by taking actions within an environment and receiving
Overview feedback based on those actions.

Dynamic Call Routing Agent Training Optimization

1. Description: RL can be used to automatically route incoming calls to 1. Description: RL can dynamically adjust the training path of call center
the most appropriate agent based on real-time data, such as customer agents based on their performance and learning speed.
urgency, agent availability, and expertise. 2. How It Works: The system receives feedback (rewards or penalties)
2. How It Works: The RL algorithm learns from past routing decisions, based on how quickly and effectively agents complete training tasks. It
How it Works analyzing which routing patterns resulted in faster resolution times or then adapts the training plan to focus on areas where the agent needs
higher customer satisfaction. Over time, it improves by maximizing improvement, ensuring more personalized and efficient learning.
these positive outcomes. 3. Use Case: Call centers can use this to optimize agent training
3. Use Case: The system can route high-value or frustrated customers to programs, ensuring each agent receives targeted training, leading to
senior agents, ensuring better handling of critical calls while minimizing improved performance and productivity
customer wait times.

Principal Component Analysis

Key 1. Agent: The decision-maker (e.g., an AI chatbot)
Characteristics 2. Environment: The setting in which the agent operates (e.g., customer interactions)
3. Actions: The choices available to the agent (e.g., responding to customer inquiries)
4. Rewards: Feedback received from the environment based on the agent's actions (e.g., positive customer feedback)

Reinforcement learning can be used to optimize chatbot responses. The chatbot learns to provide better customer service by
analyzing which responses receive positive feedback and adjusting its approach accordingly
Comparing Core Modeling Techniques

Supervised Learning Unsupervised Learning Reinforcement Learning

Definition: Definition: Definition:

Algorithms learn from labeled Algorithms identify patterns in Learning optimal actions
data to make predictions. unlabeled data. through trial and error.

Common Algorithms:
Common Algorithms: Application in Call Centers:
Decision Trees: Classify calls
based on features such as K-Means Clustering: Adaptive IVR Systems: Learning
customer history and issue type. Segmenting customers based from user interactions to improve
Random Forests: An ensemble on call behavior (e.g., call routing and reduce wait
method that improves accuracy frequency, type of inquiries). times.
by combining multiple decision
trees.
Example: Example: Example:

Predicting Customer Churn: Customer Segmentation: An IVR that adapts prompts based
Utilizing past interaction data to Identifying groups for targeted on caller responses.
identify customers at risk of marketing strategies.
disengaging
Comparing Core Modeling Techniques

Menti Activity:

Identifying Key Drivers with Supervised Learning

Clustering Techniques
 Clustering Techniques
Clustering is an unsupervised learning technique used to group a set of objects in such a way that objects in the same group (called a cluster) are more
Overview like each other than to those in other clusters. This technique is widely used in call centers to segment customers, analyze behavior patterns, and
enhance service strategies, especially for large, unstructured data to uncover hidden insights

Hierarchical DBSCAN (Density-Based

K-Means Clustering Spatial Clustering of
Clustering
A widely-used algorithm that
Applications with Noise)
Builds a tree-like structure of
partitions data into K clusters, clusters where each data point Identifies clusters based on the
where each data point belongs starts in its own cluster, and density of data points in a region,
Key Clustering to the cluster with the nearest pairs of clusters are merged as making it suitable for discovering
Algorithms mean. you move up the hierarchy. clusters of arbitrary shapes and
handling noise in the data.

Segmenting customers based on their

Grouping different customer complaints into Analyzing customer service request patterns
interaction patterns to offer personalized
Example categories to better manage escalation paths over time to identify seasonal peaks in
services or design targeted retention
and assign specialized agents. customer inquiries.
strategies
 Identifying Features for Predictive Model

Overview Feature selection is the process of selecting the most relevant variables or features for building predictive models. It reduces
overfitting, improves model performance, and helps interpret models more easily.

Feature Importance Correlation Analysis Wrapper Methods Embedded Methods

Key Concepts Quantifies the contribution Measures the relationship Feature selection occurs
Evaluate the performance of
of each feature to the between two variables, naturally within the model
a subset of features by
prediction. Commonly used helping identify redundant (e.g., Lasso Regression or
building a model.
in decision tree models. features. Decision Trees)

Predicting first call resolution (FCR), where features

Customer satisfaction score (CSAT) prediction, where
like agent experience, call duration, and customer
Application understanding which factors—such as agent empathy,
sentiment might be considered important. Feature
service time, or resolution accuracy—play the largest
selection can help identify which of these variables has
role can drive operational changes.3
the most significant impact on successful resolutions.

• Call Volume Trends: Peaks during certain hours or • Accuracy: Measure of correct predictions by the model
seasons • Precision and Recall: Important for minimizing false
• Agent Availability: Impact on customer wait times and positives/negatives, particularly in identifying high-priority calls
Sample Drivers service levels • F1 Score: Balances precision and recall, providing a single
• Interaction History: Previous issues and resolutions measure of performance.
influencing customer expectations
 Interactive Activity: Key Driver Analysis in Call Centers
Objective:
To help participants learn how to analyze a complex call center process, break it down into smaller steps, and identify key drivers that impact performance metrics such as First Call Resolution
(FCR), Customer Satisfaction (CSAT), and Average Handling Time (AHT).
Group Activity Instructions:
[Link] Groups:
1. Divide participants into groups of 3-5 members. Each group will focus on breaking down a specific call center process and identifying its key drivers.
[Link] a Dataset:
1. Give each group a dataset related to call center operations, such as:
1. Call logs (duration, type of inquiry, resolution status).
2. Customer satisfaction feedback.
3. Agent performance data (number of handled calls, resolution time).
4. Metrics like AHT, FCR, and CSAT over a specified period.
[Link] a Process:
1. Assign each group a complex call center process to analyze. For example:
1. Call Escalation Process: How customer queries are escalated from a lower tier agent to a senior agent or a specialist.
2. IVR Routing and Handling: How the IVR system routes customers to different departments or agents based on the selection.
3. Customer Complaint Resolution Process: How customer complaints are logged, tracked, and resolved.
Task:
[Link] Down the Process:
1. Each group should break the assigned process into discrete steps. For instance, in a Call Escalation Process, the steps might include:
1. Initial customer inquiry.
2. Determination of issue complexity.
3. Escalation to a higher-tier agent.
4. Resolution by the senior agent.
[Link] Key Drivers:
1. For each step, identify the key drivers that influence the process outcomes. Some examples might include:
1. Response Time: Time taken to escalate a query or route a call.
2. Agent Skill Level: Whether the agent has the expertise to resolve the issue or if it requires escalation.
3. Customer Behavior: Frustration levels or the nature of the query (technical issue, billing, etc.).
4. System Efficiency: IVR effectiveness, data retrieval time, or automation efficiency.
[Link] Impact on Metrics:
1. Groups should discuss how the identified key drivers impact key metrics:
1. First Call Resolution (FCR): Does the agent resolve the issue on the first call or does it require follow-up?
2. Average Handling Time (AHT): How long does it take to resolve the query?
3. Customer Satisfaction (CSAT): How satisfied is the customer with the overall process?
 Prompt
Engineering
Deep Dive
 Prompt Engineering
Prompt engineering is crucial in AI/ML applications as it directly influences the quality and relevance of the generated outputs. By
Overview crafting precise and contextually rich prompts, users can significantly enhance the performance of AI models, ensuring they deliver
accurate and actionable insights. This skill not only optimizes the interaction with AI systems but also empowers organizations to
leverage AI capabilities effectively, driving innovation and informed decision-making.

1 2 3
Enhancing AI Reducing Facilitating Targeted
Understanding Misinterpretation Risks Responses

Clarity & Clear and specific prompts Clear and specific prompts Specific prompts guide AI to
Specificity help AI systems grasp user help AI systems grasp user generate targeted responses
intent more effectively, intent more effectively, that align with user needs.
leading to more accurate and leading to more accurate and This targeted approach not
relevant responses. This relevant responses. This only enhances the efficiency
clarity reduces ambiguity, clarity reduces ambiguity, of AI interactions but also
allowing AI to focus on the allowing AI to focus on the fosters a more productive
essential elements of the essential elements of the dialogue between users and AI
query, thereby improving query, thereby improving systems, ultimately driving
overall output quality. overall output quality. better decision-making.
 Ineffective vs Effective Prompts
Ineffective Effective Effective
Prompt Prompt Prompt

“Tell me
Summarize the customer
feedback from the past 3
1 Clarity & Precision
about months for our online tech • A well-structured prompt should be clear and
support service, focusing on precise, eliminating ambiguity to ensure that
customer the top 3 recurring the AI understands the user's intent
feedback." complaints
• This clarity helps in generating relevant and
accurate responses tailored to specific needs

2 Contextual Relevance
• Including relevant context in prompts is
essential for guiding AI responses
• Providing background information or specific
scenarios allows the AI to generate outputs
that are not only accurate but also applicable
Clear to the user's situation
No Context Vague Specific
Objective

Context
3 Defined Objectives
• Clearly stating the desired outcome or objective
of the prompt helps the AI focus on delivering the
most pertinent information
• This approach enhances the effectiveness of the
interaction, ensuring that the responses align with
the user's goals
Strategies of Effective Prompting

1 Instruction-Based Prompts

2
3
Role-Based Prompts

Combining Approaches
Common Pitfalls of Prompting

Clarity Over Complexity Specificity is Key Iterative Refinement

Crafting prompts with clear, Regularly revising and refining

Simplifying prompts by focusing
specific details reduces prompts based on AI responses
on one key idea at a time helps AI
ambiguity, ensuring that AI can allows users to identify vague or
systems understand user intent
generate outputs that align overloaded elements, enhancing
better, leading to more accurate
closely with user expectations the effectiveness of interactions
and relevant responses tailored to
and the context of the inquiry. and improving overall output
specific needs.
quality.
Context in Prompting

1 2 3
Enhancing AI
Contextual Relevance Driving Better Outcomes
Understanding
Providing context in
Contextual information By incorporating context
prompts is essential for
reduces ambiguity and into prompts, users can
guiding AI to generate
misinterpretation, allowing significantly improve the
accurate and relevant
AI systems to grasp user effectiveness of AI
responses. Context helps
intent more effectively. This interactions. Contextual
the AI understand the
clarity results in higher prompts lead to more
specific situation, leading
quality outputs that are insightful and actionable
to outputs that are more
tailored to the specific responses, ultimately
aligned with user needs
requirements of the task at enhancing decision-
and expectations.
hand. making processes and
business strategies.
Module 2

Optimizing Call
Center Operation
BO Workflow, Intelligent Automation, Data Classification, Process
Mining, and Task Mining

Session 3
 Agenda
1 Introduction to BO Workflow in Call Centers

2 Intelligent Automation in Call Centers

3 Data Classification in Call Center Operations

4 Process Mining in Call Centers

5 Task Mining in Call Centers

Interactive Activity: Applying

6
Automation Techniques

7 Summary and Key Takeaways

8 Q&A Session

81
 Introduction to BO
Workflow and Automation
Deep Dive
 Introduction to Call Centers
Types of Call Centers Key KPIs Types of Call Centers Key KPIs

1. Inbound Call Centers: • First Call Resolution • Conversion Rate 3. Blended Call Centers: • CSAT (Customer Satisfaction) • Agent Productivity
Handling customer service, (FCR) • Calls per Agent handle both inbound and • FCR (First Call Resolution) • Uptime
• Average Handling • Contact Rate • Conversion Rate • Remote CSAT
inquiries, technical support Time (AHT) • Average Talk Time • Employee
outbound Onboarding
calls • Contact Rate • Call Quality
2. Outbound Call Centers: • Customer Satisfaction • Lead-to-Conversion • 4. Retention
Virtual Call Centers:
Strategy • Agent Utilization Rate • Cost per Call
(CSAT) Ratio • Queue Time
Sales calls, cold calling, • Abandonment Rate • Distributed teams
Employee Deboardingoperating • AHT (Average Handling Time)
follow-up, surveys • Service Level from various locations

Compliance & Quality

Operations Finance IT Human Resources
Assurance
• Call handling • Invoicing • System maintenance • Hiring • Monitoring and ensuring
• Customer interaction • Billing • Tech support • Training adherence to regulations
• Escalation processes • Budget management • CRM management • Performance • Call quality checks
• Agent monitoring • Payroll • IVR management • Employee satisfaction
 Introduction to Workflows in Call Centres
Generic Workflow Additional Elements

Call handling Ticket Omnichannel Integration

by front office creation with •Customers can interact via phone, chat, email, or social media, with
agents with customer seamless handoffs across channels.
customers details
AI-powered Self-service
Report Task
•Chatbots handle simple inquiries, reducing the need for human
generation on assignment interaction and escalating complex issues to agents.
open cases, to the
resolution appropriate Quality Assurance (QA) and Monitoring
times, KPIs team
•Supervisors monitor live calls and review recordings to ensure
compliance with policies and improve agent performance.

Real-time Analytics and Alerts

•AI-powered dashboards track KPIs like Average Handling Time (AHT)

Ticket and First Call Resolution (FCR), providing real-time insights to
closure in the Issue supervisors.
system post resolution by
confirmation BO team Workforce Management (WFM)
by customer
•Forecasting and scheduling tools ensure the right number of agents are
Update available during peak hours, optimizing resource allocation.
Communication ticket with the
with Front Office resolution or
who informs the any follow-up
customer
necessary
84
 FO Workflow Evolution

Stage 1 Stage 2 Stage 3 Stage 4

Manual Processes CRM and ERP Adoption Omnichannel Integration AI-powered Customer Experience
1990s 2000-2010 2010-2020 Early 2020s - Present

• Customer interactions • Customer Relationship • Companies integrated • AI-driven solutions enhance

were managed manually, Management (CRM) multiple communication customer experience with
Description

with call center agents systems and ERP channels—like email, predictive analytics and
relying on paper records platforms were chat, phone, and social automated bots, enabling self-
and spreadsheets to introduced to centralize media—into unified service and faster issue
track customer data and customer data and systems to offer resolution.
queries. streamline workflows. seamless customer
experiences.
• A retail company uses • A telecom company deploys
• A financial services Salesforce to track • An airline provides real- AI chatbots with Genesys to
company maintains
Use Cases

customer interactions, time support via chat, handle routine inquiries and
customer records and while ERP platforms email, and phone, with improve service availability
handles service requests manage sales and order customer interactions 24/7.
manually via phone calls. data. logged in Zendesk.

Paper files Salesforce SAP ERP Zendesk Genesys

Spreadsheets Microsoft Dynamics Freshdesk

 BO Workflow Evolution

Stage 1 Stage 2 Stage 3 Stage 4

Basic Digitization Advanced Workflow Automation Cognitive Automation Intelligent Automation

Early 2000s 2010-2015 2016-2020 Early 2020s - Present

• Companies used early • RPA bots automated • AI-enabled workflows • Combining RPA and AI,
ERP systems to digitize repetitive, rule-based handled unstructured organizations automate
Description

procurement, invoicing, tasks like extracting and data, improving entire processes, from
and approval workflows, validating invoice data. processes like document order entry to payment
replacing manual classification and claims reconciliation and
processes processing beyond

• A bank uses UiPath to • An insurance company • A telecom company

• A retail company automate invoice uses Kofax for AI-driven automates customer
Use Cases

automates PR/PO matching and exception claims management order management with
creation and invoice handling in accounts Blue Prism and AI
management with SAP payable integration
ERP

SAP ERP Oracle ERP Automation Anywhere IBM Watson Kofax

Ui Path
Intelligent Automation
Deep Dive
 Intelligent Automation
Intelligent Automation (IA) refers to the use of advanced technologies, particularly Robotic Process Automation (RPA) and Artificial Intelligence (AI), to automate business processes,
handle high-volume tasks, and optimize decision-making. Unlike traditional automation, which focuses on repetitive rule-based tasks, IA incorporates cognitive abilities like learning,
reasoning, and self-correction through AI

Key Benefits Example Use Cases of Intelligent Automation

• Efficiency Gains
• Reduction in Repetitive Tasks • AI Chatbots and Virtual Assistants
• Faster Call Resolution • AI-driven chatbots are commonly used in call centers to handle
• Cost Reduction common queries like FAQs or simple service requests
• Lower Operational Costs • Predictive Call Routing
• Improved Resource Allocation • AI-based systems can predict customer needs based on historical
• Scalability data and route calls to the most appropriate agent or department
• Handling Increased Call Volumes • Automated After-Call Work (ACW)
Label this

• Improved Customer Experience • RPA bots can automatically log call details, update customer
information, and summarize the call interaction in the CRM system

Challenges and Considerations

• Integration with Legacy Systems

• Many call centers use legacy systems that may not easily integrate
with modern automation tools
• Change Management and Employee Buy-In
• Automation can cause resistance among employees who may fear
job loss or feel displaced by technology
• Continuous Monitoring and Optimization
• Automated systems require continuous monitoring to ensure that
they perform optimally and adapt to evolving business needs
Case Study - IA

In this section, we’ll deep dive into a real-world case study where automation, specifically Robotic Process
Automation (RPA) and Artificial Intelligence (AI), transformed a call center’s operations. The case focuses on
a global telecommunications company that sought to improve its customer service efficiency while
reducing operational costs.

Overview:
A telecommunications company managing over 10,000 customer support inquiries per day across different channels:
phone, email, and chat

High Average Handling Time (AHT): Manual Post-Call Work:

The company’s AHT was 7.5 minutes, After each call, agents spent an
Challenge leading to long customer wait times additional 3 minutes updating CRM
systems and creating call summaries

Escalation Rate: Agent Utilization:

25% of calls were being escalated Due to the volume of post-call
due to improper routing, leading to administrative work, agent utilization
inefficiencies in the overall workflow was low (around 60%)
 Develop IA Model
Automating After-Call Work (ACW) Desired Outcomes

• Implemented RPA bots to automate post-call 1 Reduction in AHT

tasks, such as logging call details, updating
customer information in the CRM, and generating
follow-up tasks 2 Increased Agent Efficiency

3 Escalation Rate Improvement

Intelligent Automation 4 Cost Savings

AI-Driven Call Routing

• Used AI-based Natural Language Processing

(NLP) to automatically categorize and route calls 5 Customer Satisfaction
based on customer intent, reducing misrouted
calls and escalation rates
RPA 6 24*7 Service Availability

AI Chatbots for FAQs

• Deployed AI-powered chatbots to handle

common customer queries, such as balance
inquiries, bill payments, and troubleshooting,
reducing the need for human intervention in these
areas
AI Chatbots

90
Success Story: Intelligent Automation
AT&T Inc. is a global leader in telecommunications, media, and technology services. Founded in 1983 and headquartered in
Dallas, Texas, AT&T is one of the largest telecom companies in the world, serving millions of customers across the U.S. and
globally

High Post-Call Low Agent Inconsistent Call High Operational

Challenge Processing Times Utilization Handling Efficiency Costs

Phase 1: Process Mapping and Analysis Phase 2: RPA Bots & AI Deployment Phase 3: Continuous Improvement

• AT&T monitored performance

• Conducted an in-depth analysis to identify • RPA bots were introduced to automate
metrics such as Average Handling
tasks suitable for automation, such as after-call tasks and report generation,
Time (AHT), First Call Resolution
repetitive call logging, CRM updates, and reducing the manual workload for agents.
(FCR), and agent utilization to
Approach responding to frequently asked customer
inquiries.
• AI-driven tools, such as Natural Language
Processing (NLP) and machine learning
measure the effectiveness of
automation.
• Process mining techniques were employed models, were integrated into the call routing
• Regular feedback from agents and
to map inefficiencies in the existing system to ensure more accurate call
customers was gathered to
workflows and develop an automation distribution and reduce unnecessary
continually refine and improve
roadmap escalations
automated workflows.

Annual Savings Reduction in Increase in

through Increase in post call
Impact $2M efficiency 25% Agent Utilization 40% processing 30% customer
inquiries
improvement times handled
Best Practices for IA
• High Volume and Repetitive Tasks: Call summaries, Data entry into CRM systems, Call
routing based on predefined criteria
Assess Automation Needs • Error-Prone Tasks: Processes like manual data entry or repetitive customer interactions (e.g.,
balance inquiries) are prone to human error and should be considered for automation

• Call Routing: AI-driven IVR (Interactive Voice Response) systems can route calls based on
customer intent, ensuring customers are connected to the right agent.
Prioritize Customer Facing Processes • Self-Service Options: Implement AI chatbots to handle basic customer inquiries, such as
checking account balances, resetting passwords, or providing order status updates

• Testing and Adjustments: You can assess how well the automation tools perform and make
necessary adjustments based on feedback from agents and customers
Pilot the Automation Program • Minimizing Disruption: Starting small allows the center to manage potential issues (such as
integration problems) without disrupting overall operations

• Customer Relationship Management (CRM) Systems: Ensure the automated workflows can
Integration with Existing Systems interact with the CRM to pull customer data or log interactions in real-time
• Call Routing and IVR Systems: Automation should integrate with these systems to improve
routing accuracy and reduce call handling times

• Involve Employees Early: Engage call center agents and supervisors early in the automation
Change Management & Employee process
Training • Provide Comprehensive Training: Equip agents with the knowledge and skills they need to
work with automation tools

• Expand Automation to New Processes: Once initial successes are observed, move to
Continuous Improvement automate other back-office functions
• Refine Workflows: Continuously monitor workflows to find new optimization opportunities
Data Classification
Deep Dive
 Data Classification in Call Center Operations
Data classification is the process of organizing data into categories or groups so that it can be more efficiently used, stored, and secured

Customer Interaction Operational Historical

Types of Data
Information Data Data Data

Manual Classification Rule-Based Classification AI-Based Classification Metadata Tagging

Techniques for Data AI and machine learning

Data is classified Assigning metadata to
Classification Agents or supervisors algorithms are used to
automatically based on a set different pieces of
manually categorize data classify data dynamically
of predefined rules or information to organize and
based on predefined criteria based on patterns in the
conditions classify them efficiently
data

A machine learning model that

In a call center, each
If a call contains the keyword categorizes customer
Use Cases this method is time-consuming interaction might be tagged
“refund,” the system might complaints based on
and prone to errors, especially with metadata like call type,
automatically categorize it sentiment analysis, identifying
in large-scale operations. duration, and customer
under “billing inquiry” whether an issue is technical
satisfaction rating
or service-related

Efficient Data Improved Decision Compliance and

Benefits
Retrieval Making Security
Process Mining
Deep Dive
 Process Mining
Process mining is a technique that extracts knowledge from event logs available in call center systems to understand how processes are carried out,
versus how they were intended to be

Discovery Conformance Enhancement

Types of Process AI and machine learning

Data is classified
Agents or supervisors algorithms are used to
Mining manually categorize data
automatically based on a set
classify data dynamically
of predefined rules or
based on predefined criteria based on patterns in the
conditions
data

A telecommunications
A call center might use A financial services call center company could use
Use Cases discovery techniques to may use conformance mining enhancement mining to
understand the actual flow of to ensure that agents follow identify which steps in a
call escalations and uncover the correct escalation path customer inquiry resolution
inefficiencies in how calls are when handling sensitive process are causing the most
routed between agents customer data delays and tweak their
workflows accordingly

Benefits Improved Efficiency Reduced AHT and Wait Times Compliance Monitoring

Key Vendors
 Process Mining in Call Centers
Mining Preparation Process Mining Analysis Optimization & Improvement
Steps in Process 5. improvements like automating call
3. Process mining creates workflow
Mining 1. Data is gathered from systems like models and compares them to KPIs routing or enhancing agent training are
CRM, IVR, or ticketing platforms in the like AHT, FCR, and CSAT implemented to streamline workflows
call center 4. Managers gain insights into 6. The changes are continuously
2. This data is structured into event logs, inefficiencies, such as call routing monitored to ensure improvements,
containing key details like case ID, issues, for process optimization with regular checks to address any new
activities, and timestamps to track the inefficiencies
process flow

1 6 5

Process Flow

3 4
2
Success Story: Process Mining
Vodafone Netherlands, a leading telecommunications provider, serving millions of customers across the country with mobile,
internet, and TV services

High Variability in
Inconsistent Order-to- Low Customer
Challenge Activation Process
Long Activation Times
Satisfaction (CSAT)
Regional Service
Efficiency

Phase 1: Process Mining & Assessment Phase 2: Optimization & Standardization Phase 3: Continuous Improvement

• Vodafone Netherlands implemented • Vodafone standardized its order-to- • The company implemented a
Celonis process mining to analyze their activation process based on insights from continuous monitoring system to
order-to-activation process across process mining data, reducing unnecessary track process efficiency in real-
regions. steps and automating certain tasks
time
Approach • Discovered significant variability in
activation times between different regions,
• Introduced process enhancements across
all regions to align with the most efficient
• Provided targeted training to
regional teams to ensure process
leading to delays and customer complaints. practices identified by the process mining adherence and used KPI
• Created a detailed roadmap for analysis dashboards to identify further
standardizing the activation process and optimization opportunities
reducing delays across regions.

Reduction in Increase in Improvement

Impact 50% order activation
delays
25% Customer
Satisfaction
30% in process
transparency
Task Mining
Deep Dive
 Task Mining
A technique that captures and analyzes the detailed tasks performed by employees on their workstations. It focuses on task-level data, such as mouse
clicks, keyboard entries, and application use, to identify inefficiencies and areas for automation

Data Collection Task Analysis Task Classification Opportunity Identification

Task mining tools categorize
Steps of Task Software agents are installed
The system compiles the tasks based on patterns, such Data is classified
on employees' systems to
Mining monitor how they interact with
captured data into sequences, as identifying repetitive tasks automatically based on a
showing how users perform a like updating customer set of predefined rules or
applications during the
task step-by-step records, generating reports, or conditions
completion of their daily tasks
sending follow-up emails

Data captured:
In a call center, task mining Agents spending excessive
• Mouse clicks
Use Cases can track how agents handle Classifying and labeling tasks time manually logging post-
• Keystrokes
customer inquiries, enter data helps managers prioritize call notes into the CRM
• Transitions between
into the CRM, and follow up which activities to automate system can be automated
different software
with emails or tickets using RPA bots.
applications

Granular Insights into Identifying Automation Improving Employee

Benefits
Workflows Opportunities Productivity

Key Vendors
Thank You