- Chapter I: Introduction & Definitions
- Chapter II: Enhancing Agent Productivity and Efficiency
- Chapter III: Improving Customer Experience and Engagement
- Chapter IV: Automated Quality Management, Agent Performance & Training
- Chapter V: AI-Powered Knowledge Management
- Chapter VI Data Analysis and Insights
- Chapter VII What’s Next For Contact Center AI?
- Chapter VIII FAQ: Implementing Generative AI In Contact Centers
Automating Post-Call Summary
When done manually, post-call wrap-up time can take, on average, anywhere from one to five minutes. Now multiply that by the 50 calls a contact center agent handles on a daily basis, and you are looking at 50 to 250 minutes of administrative time. Generative AI can streamline a lot of these post-call processes by automatically generating call summaries and disposition codes for CRM systems. This automation not only saves time but also standardizes summaries for future reference.
Examples: Five9 AI Summaries, MiaRec Auto Call Summary, Talkdesk Agent Assist, Salesforce Call Summaries
Image: Screenshot showing Five9 AI Summaries.
Image: Screenshot showing Talkdesk AI generated summary.
Image: Screenshot showing Salesforce Call Summaries powered by Einstein.
Automating Note Taking
Very similar to automatic post-call summary is Generative AI’s capability to automatically jot down important information DURING a customer call. This feature enables agents to better focus on the conversation as they no longer have to multitask between listening and note-taking. This tremendously improves call handling efficiency and helps prevent agent burn-out.
Example: Observe AI Auto Summary, Uniphore Agent Assist’s Summary & Disposition
Image: Screenshot of Observe AI Agent Assist
Example: Uniphore Agent Assist’s Summary & Disposition
Auto-Generating Customer Replies
Because Generative AI can understand intent, it can first read and understand a customer’s query and then try to automatically generate a helpful response by sifting through relevant sources like CRM systems, knowledge bases, customer databases, web links, and such. However, it is important that agents review and edit the responses before forwarding them to ensure their accuracy.
Examples: Salesforce, HubSpot, Talkdesk
Image: Example of Salesforce automated chat
Image: Example of the HubSpot chatbot builder
Image: Example of Talkdesk AI writes responses to customers
Auto-Completing Agents’ Responses
Solutions like Zendesk's "expanding agent replies" use Generative AI to complete responses for agents. Instead of just auto-completing his sentences like Google AutoComplete, these solutions will deploy AI to access various knowledge sources to finish an agent’s response, saving significant time in customer communication.
Example: Zendesk's "expanding agent replies"
Image: Screenshot of Zendesk expanding agent reply
Real-time Language Transcription and Translation
AI can provide real-time or near real-time speech-to-text transcriptions. You can also use Generative AI to instantly translate customer queries, enabling support in multiple languages and reducing language barriers. Real-time translations of customer conversation is huge as it allows an organization to broaden their customer base and enhance their global support capabilities. This is relatively common in text-based communication channels, but is starting to emerge in voice-based channels by augmenting your agents’ voices when translating.
Example: Parloa
Image: How Parloa real-time translation works
Streamlining Call Transfers and Escalations
Transferring and escalating calls is often time-consuming as it requires a knowledge transfer to update the agent overtaking the conversation. Generative AI tools can reduce wait times by summarizing conversations and transferring them to the next available support agent or supervisor. This reduces the need for customers to repeat their issues, streamlining the escalation process.
Example: Verint Interaction Transfer Bot
Image: Amazon Connect Catch up Summary during call transfer
Step-by-step Guidance of Agents in Real-time
An AI assistant can recommend the next best action in real-time based on conversation.
Example: Uniphore’s U-Assist, Five9 Agent Assist with Real-Time Guidance
Image: Uniphore’s U-Assist agent assisted real time coaching
Image: Five9's Agent Assist giving real-time guidance
Searching Knowledge Base for an Answer to Customer’s Questions
AI-based assistants can automatically find answers to customer’s questions, eliminating frustrating and tedious manual searches. This is leading to an increase in first-call resolution and a decrease in agent errors. It can also improve customer experiences as the agent becomes more in tune with the customer’s concerns and has solutions at hand to resolve the problem right away, without putting the customer on hold or having to transfer them.
Example: Uniphore Agent Assist, Genesys Agent Assist
Image: Uniphore's Agent Assist with real-time coaching
Image: Genesys chatbot using knowledge base answers
Post-Call Agent Coaching
You can also use Generative AI to offer automated performance summaries after a call has ended. These summaries can be used to provide feedback and develop improvement strategies for agents, thus enhancing their performance and the overall customer experience.
Example: Manager Assist for Amazon Connect, MiaRec Assist
Image: Amazon Connect’s Manager Assist
Image: MiaRec’s Agent Assist auto call summary example
Automating processes
Where an AI Assistant accurately predicts the next best action(s) for the interaction, the process automation robot can trigger backend tasks on behalf of the agent, saving valuable time and eliminating manual errors.
Example: Uniphore robotic process automation
Image: Uniphore's robotic triggering of tasks in the backend
Generative AI plays a crucial role in improving customer experience and engagement in contact centers by providing personalized, efficient, and empathetic interactions.
Empower Self-Service through AI-powered Chatbots and Virtual Assistants
One of the maybe most straight-forward applications of Generative AI is the implementation of chatbots and virtual assistants for self-service. They enable customers to independently address common queries by typing in their question or concern into a chatbot or conversational interface. This allows support teams to focus on more complex issues, reducing hold times, lowering Average Handling Time (AHT), and increasing customer satisfaction.
Example: Amazon Lex Chatbot
Image: Lex chatbot on mobile
Example: HubSpot Chatbot, TalkDesk, Vonage
Image: TalkDesk chatbot with dashboard
Improving Self-Service Calling Experience
AI-powered Conversational Interactive Voice Response (IVR) uses Natural Language Processing (NLP) to ingest, analyze, and understand customer requests and then provide an automated response without the help of a live agent. It can create personalized experiences and, if needed, will direct customers to the appropriate self-service or agent channels based on their stated needs.
Example: GetVOIP, Teneo.ai, Five9 Intelligent Virtual Agent
Image: GetVOIP IVR call flow
Understanding Customer Sentiment
Another common use case for Conversational and Generative AI is analyzing customers' sentiments, moods, and emotions during interactions. This can be used live to gauge appropriate reactions, or be used post-call for quality management. For example, measuring how well an agent can navigate difficult conversations (angry customer = negative sentiment score) and bring them to a good outcome (happy customer = positive sentiment score) can give you real-world insights into an agent’s people skills and emotional intelligence. This helps to foresee and prevent escalations, improve satisfaction, and reduce churn rates.
Example: MiaRec Sentiment Analysis
Image: MiaRec's dashboard on sentiment analysis
Identify Potential Growth Opportunities
By analyzing conversation data, Generative AI can identify potential issues or growth/upsell opportunities (e.g., increased calls about shipping problems, calls requesting a feature that is not available) proactively. Essentially, AI mines your customer interactions to determine the interactive drivers, and surfaces these insights as trends or shifts in distribution.
Example: MiaRec Topic Analytics, Five9 Insights using OpenAI
Image: Miarec’s topic analytics dashboard
Automated Agent Evaluation and Call Scoring
Rather than manually evaluating a small fraction of calls at random, Generative AI can be used to automatically score 100% of your calls, providing a much more accurate representation of agent performance over time. These so-called Auto QA tools free up a tremendous amount of time every week that can be used in higher-value activities, such as focusing on calls flagged for follow-up or review.
Example: MiaRec Auto Call Scoring
MiaRec’s Auto Score Card
Adding Context to Automated Quality Scoring
For some questions like “Did the agent say their name?” the answer is simply a yes or a no, while answering questions like “Was the agent polite?” requires more context. While Generative AI can be used to automatically evaluate the call and gauge the level of politeness, it can also provide context as to why the answer might be “no” by including a quick statement about how the agent responded rudely to the customer’s request for a refund. This gives the supervisor enough context to dig deeper if needed.
Example: MiaRec Auto Call Scoring
Image: MiaRec's auto call scoring adding context
Identifying Agent Training Opportunities
Similarly to post-call coaching, Generative AI can identify opportunities for improvement and make individualized and unbiased suggestions for additional training based on the call evaluation to enhance the agent’s performance.
Example: MiaRec Auto Call Scoring
Image: MiaRec's auto call scoring identifying training opportunities
Alerting Supervisors to Issues With Their Agents
As a contact center supervisor, you always want to know if something extraordinary happens with your agents. Maybe your agent is struggling with a personal crisis and comes to work sleep-deprived and in a foul mood, which he takes out on the customers he or she deals with. Generative AI can identify if swear words are being used, if the mood in the call suddenly goes icy (sentiment analysis), or if the customer gets angry. As a supervisor, you will be notified and can intervene as soon as possible.
Example: NICE Enlighten Actions, MiaRec Auto QA + Notifications
Image: NICE Enlighten Actions dashboard
Image: Example of an automated agent scorecard from MiaRec
Automating Compliance Management (Data Redaction)
Contact centers process an enormous amount of Personal Identifiable Information (PII) and payment card numbers on a daily basis. Those get stored in your call recordings and the associated transcripts. In most cases, you are obligated by regulatory compliance to redact them. Machine Learning and Named Entity Recognition (NER) can reliably identify and redact sensitive information from your calls and transcripts, increasing your compliance and decreasing your liability should you suffer a data breach.
Example: MiaRec Auto Redaction
Image: MiaRec’s auto redaction in a call transcript
Supporting content
- MiaRec Auto Score Card
- Auto Redaction
- 7 Misconceptions about AI-Based Auto QM in Contact Centers
- New Automated QA Feature Update Helps You Improve Auto Score Card Accuracy
- Introducing MiaRec's Auto Score Card: Evaluate Agent Performance On 100% Of Your Calls
- Auto QA ROI Calculator
- MiaRec Takes Auto Data Redaction To New Heights With Machine Learning
The long-term goal of utilizing Generative AI in your contact center is to automate as many calls and inquiries as possible by providing the answer in an external- or internal-facing knowledge base and having an AI-based chatbot or virtual assistant answer them. The following five use cases help you get there.
Spotting Gaps In the Knowledge Base
Generative AI can identify when there are gaps in a contact center's knowledge base, flagging them to management for the creation of new knowledge articles or improving existing articles based on the collected feedback.
Example: TalkDesk Knowledge Management
Image: TalkDesk Knowledge Management spotting gaps
Other Examples: Google CCAI Generative FAQ, Twilio CustomerAI, Verint Knowledge Creation and Suggestion Bots
Generating Knowledge Base Articles
After it has identified gaps, Generative AI can also be used to fill these gaps by creating new knowledge articles. It can do this by understanding the customer intent and analyzing how agents and supervisors have successfully handled similar queries in the past. But as always, be the human in the loop and review and edit anything before it gets published.
Example: Twilio CustomerAI, Verint Knowledge Creation Bot
Image: AI generated knowledge base
Generating FAQs from Customer Conversations
This involves AI tools analyzing customer conversations to generate FAQs, helping contact centers better understand and address common customer queries.
Example: Google CCAI Generative FAQ
Chat-Like Information Discovery
Rather than having to read help articles in the knowledge base with the customer waiting on hold, the agent can use AI-enhanced search. Essentially, an agent writes a quick query and Generative AI augmented search capabilities will auto-generate the most probable answer based on the internal knowledge base. This not only improves access to information in knowledge bases, but also makes knowledge discovery much faster and easier.
Example: Uniphore Agent Assist with Knowledge Base search
Image: Uniphore's Agent Assist recommendations
Generative AI can provide you with powerful insights and help you discover trends in your consumer behavior, preferences, and intent.
Call and Trend Analytics
AI can provide more comprehensive call analytics and identify trends more accurately than a human customer support manager because it tracks customer sentiment, tone, and personality, and it has a much more complete picture of what is going on across all calls. It is often utilized for in-depth analytics on various aspects of calls, such as call times and first resolution rates.
Example: MiaRec Sentiment Trending Dashboard
Image: Screenshot of MiaRec's Sentiment Trend dashboard
Automatic Keyword Spotting
One of the easiest but nevertheless powerful applications of generative AI is to automatically spot certain keywords in call transcripts to identify calls that mention competitors, brand or product names, specific features, marketing campaigns or similar information. This can offer valuable insights into customer preferences and market trends.
Example: MiaRec Voice Analytics
Image: A sample sentiment score from MiaRec's Voice Analytics
Discover Call Reasons and Extract Customer FAQs
The ability of Generative AI to uncover customer intent and categorize primary reasons why your customers contact your business is very helpful in getting better insight into call reasons. This helps in understanding demand drivers and prioritizing actions to improve contact center operations. In addition, you can extract frequently asked questions and answer them in your knowledge base.
Example: MiaRec Voice Analytics, Google Cloud’s Generative FAQ for CCAI Insights
Image: Call reason trend report in MiaRec
Extracting Insights from Customer Feedback
Generative AI solutions can ingest and analyze large amounts of customer feedback from various sources (such as call transcripts, surveys, and online reviews), distill them into structured, actionable insights, and enable continuous monitoring and quick responses to emerging issues.
Example: Verint Insights Bot
Identifying What Makes Conversations Successful & Define Resolution Paths
Generative AI can assess thousands of customer call transcripts and assess which ones are the most successful conversations. Then it can identify what worked and what didn’t, resulting in ideal paths for call resolution. This is invaluable insight that informs agent coaching and scorecard creation initiatives.
Example: Cresta Flow Modelling
Measuring The Performance of Marketing Campaigns
Generative AI tools can track and analyze customer interactions to assess the effectiveness of marketing campaigns, providing insights into customer engagement and responses to marketing efforts. In the simplest case, the marketing campaign has a specific coupon code the customer needs to mention when placing an order or booking a service that is then recognized as a keyword to be tracked. Or you can measure an uptake in interest in a specific product or service that has been recently promoted.
Example: MiaRec Voice Analytics
Image: Measuring marketing campaigns with MiaRec's call summary
Insights Made Clear: Introducing Upgrades to MiaRec's Topic Analysis
Using AI-Extracted Insights from Contact Center Call Recordings to Reduce Customer Churn
