For the past decade, conversational intelligence (CI) has transformed the contact center from an opaque river of interactions to a vibrant source of business analytics. The basis of conversational intelligence is the ability to analyze and understand the context of a conversation to build trust between customers and organizations.

Conversational intelligence combines real-time and post-interaction analytics that fuel real-time intelligence through bots, virtual agents and agent-assistive tools. Artificial intelligence (AI) plays a crucial role in both analyzing large amounts of data in and beyond an interaction to provide useful insights and continuous optimization to improve customer experiences and business results.

How New CI Tools Will Drive a Better Customer Experience

With the maturity of AI-based tools like voice transcription, entity and intent recognition, and sentiment analysis, among other technologies, supervisors and contact center managers gain a clear idea of what’s currently happening in interactions and can forecast trends. However, with the convergence of large language models and generative AI, as well as the sharp reduction in cost, increase in speed and improved reliability of existing tools, CI is set to again transform the contact center.

It will take the contact center from a place where analytics are acquired to one where customer concerns are resolved even before an agent needs to intervene.

What is Conversational Intelligence? Laying the Foundation

Conversational intelligence is built on layers – each of which unlocks new capabilities that enable the supervisor or contact center manager to answer key questions:

  • What is happening in the contact center now?
  • Why are people calling?
  • Are customers happy with the outcome of the interactions?

While live call monitoring has been around since the switchboard was invented, the creation of call recording allows supervisors to free themselves of the non-scalable task of listening in live on calls. They could review calls when it was convenient for them.

The next evolution and building block toward true conversational intelligence occurred with automated voice transcription. It was in the past five years that voice transcription achieved a level of accuracy and reliability that made it possible to read transcribed text and understand what happened in a conversation. This meant that supervisors wouldn’t have to listen to recordings. They could scan over the transcripts of conversations to glean what happened.

By unlocking the text of interactions, many new technologies could now be applied to achieve even better understanding. Text can be searched; searches can be automated; phrases can quickly be spotted to identify topics. And these topics can be compared over time to identify trends and surface information that might be important for supervisors to be aware of during their shifts.

On top of basic search, natural language understanding through entity and intent recognition can identify the reasons why customers call. New tools based on AI can also analyze both sides of the conversation to look at whether the customer expressed positive or negative sentiments and if the agent was empathetic or unhelpful in response.

By looking at the acoustics of an interaction, supervisors can also extract information that would require many hours of listening and transcript reading to gather.

  • Did the agent talk over the customer?
  • Were there awkward pauses?
  • Was one side of the conversation speaking for much longer than the other?

Putting It Together: Conversational Intelligence Now

The full power of conversational intelligence to date was unlocked when the information acquired through these different tools was fused to create new understandings of individual interactions, agent performance and overall trends. Sentiment analysis can be viewed through individual events, overall scores for an interaction or a trend of whether the call started off with negative sentiment then transformed to positive sentiment, vice versa, or whether it had no movement.

By rolling up the sentiment and empathy scores for interactions, post-call survey results and other metrics, supervisors can see how agents perform both individually and relative to their peers. The same rolling up of metrics can be done at the queue level to determine which queues are performing better. This allows supervisors and contact center managers to optimize performance around KPIs. More information can be extracted by combining data sources to categorize interactions.

  • Did the agent or customer mention certain topics in proximity to each other?
  • Was a topic said at the beginning or end of a conversation as the sentiment was trending negative?
  • Did the agent interrupt the customer when negative sentiment was present?

By creating buckets of interactions that meet specific criteria, supervisors can accumulate a treasure trove of examples for teaching agents what’s more or less effective when handling interactions.

The Next Level for CI Tools

The building blocks to generating actionable conversational intelligence are present with the current generation of conversation intelligence tools. But the next generation of tools will focus on surfacing information to agents and supervisors right at the time they need it — and in a clear concise manner. These tools will also propose solutions and next steps and provide reasoning behind their suggestions.

Many of the current generation of CI tools use rigid rules engines to make determinations of whether data is actionable. The next generation feeds information into large language models to formulate an analysis.

As a first step, generative AI tools can create summaries of individual legs of interactions, entire conversations or even high-level summaries of multiple interactions that have occurred by a single agent. The tools can also summarize an entire queue over a specified period for a more comprehensive view of performance.

Generative AI tools can also be used to uncover insights into what happened in an interaction.

  • What was the reason for the call?
  • What was the resolution?
  • What were the action items that the agent needs to undertake?
  • If there was a strong positive or negative sentiment score during the interaction, what was the cause?

This new abstraction saves the time needed to stitch together topics that were spotted in interactions or sentiment scores to get to the true reason a customer is interacting with an agent. These insights can be used to improve agent performance, bot knowledge, agent copilots and virtual agents to improve overall contact center performance.

The next generation of tools will do a better job of surfacing critical information as it’s needed. Supervisors will not only be able to roll up analytics but will also be able to quickly dive into the numbers to look at examples from individual interactions. They can then use these examples for training or to illustrate why certain trends are happening.

Supervisors will also be able to see — in real time — what’s happening in any conversation in the contact center. They can view live transcription of interactions and tap into the topics being discussed at any moment on a queue, and the corresponding sentiment of customers and helpfulness of agents. This creates a chance for either AI-based services or supervisors to intervene if an interaction isn’t going well.

Beyond providing more flexibility and speed in monitoring interactions, CI tools will start to formulate responses from agents to customers that won’t only resolve individual reasons for contact but will also serve the larger business goals of the contact center. At the supervisor level, this could include natural language suggestions on what a supervisor can do in the moment to address problem areas, maintain existing or attain new performance records, or even to commend exceptional performance and reward great agents.

Paving the Way

Deploying the next generation of conversational intelligence isn’t without risk. Every generative AI-based system must be safeguarded against hallucinations, prompt injection attacks and biases in training. Staff must be trained to work alongside the tools. And you need to ensure there’s a balance of cooperation and skepticism with the results. Great care must also be provided to ensure that training these systems doesn’t compromise privacy or ethics.

The upside of the next generation of tools is that they’ll lead to much happier customers. And they’ll allow employees to focus on the core parts of the interaction, rather than data entry or reporting.

Customers with relatively simple issues will have their problems resolved before they get in a queue, leaving agents to deal with complex, more satisfying interactions.

With Genesys Cloud Speech and Text Analytics you can quickly analyze text and voice interactions – on any channel. Learn more now. And get a roadmap for adopting AI capabilities that deliver ROI now and set the foundation for great long-term value.