Contact Center Pipeline September 2025(clone) | Page 28

These sophisticated AI agents, powered by GenAI and conversational AI, can listen to each caller’ s needs and then respond with vital information or instructions the caller is seeking. Human agents can even manage these AI agents remotely, monitoring calls for quality. Or, if the need is more complex, the AI agent can answer, listen, collect relevant information, and then hand off the call to a human agent for completion.

The leap forward in GenAI and conversational AI presents an opportunity for contact centers to build a new conversational AI playbook. AI agents can help understand, contextualize, and anticipate customer needs, enabling more personalized and consistent interactions.

HOW AI AGENTS“ HEAR” THE CUSTOMERS

In most contact centers, audio-to-audio communication between a customer and an AI agent typically occurs over several stages( see FIGURE 1). These stages are:

1. A person speaks to the AI agent.

2. The AI agent converts the caller’ s audio into text( speech-to-text or STT).

3. The AI agent recognizes speech activity through voice activity detection( VAD) and contextual endof-speech detection.

4. The AI agent thinks and reasons about the best response, using either deterministic( rule-based), non-deterministic( GenAI / large language model [ LLM ]), or hybrid approaches.

5. The AI agent reads the text, generates its response in text form, and then converts that text into audio( text-to-speech or TTS).

6. The AI agent’ s audio is sent back to the person.

Although these six steps occur in rapid succession, each conversion— from audio to text and back— can introduce slight latency. Streamlining these steps is critical to achieving a fluid, natural conversation between the customer and the AI agent.

As call centers and customer support teams increasingly embrace real-time, voice-based AI capabilities, solving this latency and reducing the number of steps in the communication chain will be crucial for customer satisfaction and broader adoption of AI communication.

SOLVING THE LATENCY PROBLEM

This is where advanced GenAI becomes a game-changer. For example, OpenAI’ s Realtime API seeks to improve this chain by eliminating the need for interim text transcriptions between the customers and the AI agents, reducing response times, and creating more seamless conversations.

With capabilities like embedded VAD and end-of-speech recognition, interactions are more attuned to natural speech, further narrowing the gap between AI and human agents. Such solutions typically have a built-in barge-in capability through the Realtime API, which is also a benefit because it allows the caller to interrupt the AI agent if the conversation gets off track( see FIGURE 2).

Audio-to-audio exchanges also have the potential to capture additional dimensions of a conversation that aren’ t otherwise encompassed in text. Think about the missed nuances between a text message and a phone call with a friend. Tone, emotion, emphasis, speed, and other human elements of spoken conversation now have the opportunity to influence and better personify exchanges with AI agents.

THE HUMAN IN THE LOOP

Even as this technology matures, human agents will always be needed to solve more complex problems. In those cases, AI can also make the transition process more seamless by providing a better way to define escalation criteria for AI agents through natural language briefings.

FIGURE 2:

SOURCE: PARLOA

28 CONTACT CENTER PIPELINE