Press 0 for the Future: Voice AI Agents Set to Transform Contact Centers in 2025

Brought to you by MiaRec Founder and CEO, Gennadiy Bezkorovayniy .

The largest cost in a contact center is human labor—specifically, the wages and benefits of call center agents, supervisors, and other staff involved in customer interactions.

Over the last few years, companies have been trying to increase the efficiency of contact centers by either replacing agents with technology (IVR, self-service portals, chatbots) or empowering agents to do more through tools like agent assist, better quality monitoring, and personalized training.

Almost all companies today are placing IVR systems and chatbots at the frontline of their customer service to act as gatekeepers for their valuable human agents. But let's be honest, customers hate IVRs and chatbots. When faced with an IVR system, most callers simply press "0" to reach a real human who can understand their unique problem. The majority of IVR systems are just scripted decision trees: "Press 1 to do X, press 2 to do Y, press 0 to talk to a representative." Customers are neither dumb nor lazy. If their issue were as simple as "X" or "Y," they would figure it out themselves or search online for a solution. When they call a contact center, their problem is often more complex than what can be addressed by an IVR system.

Chatbots offer more powerful capabilities than IVR systems. Early chatbots were essentially text-based IVRs with the same limitations: scripted and narrow in scope. With the mass adoption of Large Language Models (LLMs), chatbots have become smarter and more capable. They can now handle medium-complexity tasks, find answers to customer queries in internal knowledge bases, and even perform simple actions on behalf of customers, like rescheduling appointments or submitting refund claims.

Chatbots may seem like the go-to solution, but they have one core weakness: they rely on a text-based interface. For computers, this interface is efficient. For humans, however, voice is a much more natural form of communication. With chatbots, we force humans to adapt their communication to suit computers. But what if we flipped the script and adapted computers to suit humans instead?

How about voice bots? A system that uses a human-friendly voice interface but is far more capable than outdated IVR systems.

Does this sound too good to be true???

A year ago, it was. Early voice bots were smart but not user-friendly, mainly because:

1. They were too slow. The process involved listening to the caller, transcribing the voice using speech-to-text models, processing the text with an LLM, searching internal knowledge, generating a response, converting the response back into voice using text-to-speech models, and then playing the response to the caller. While a few seconds’ delay is acceptable in text-based chats, it’s frustrating in voice conversations.
2. They couldn’t be interrupted. Once a voice bot started delivering its response, the user had to listen to the entire message—even if it wasn’t exactly what they needed or if they wanted to correct the bot mid-response.

This year (2024), we’ve finally seen the emergence of products that address these fundamental issues. Modern voice bots are faster, allow interruptions, and retain the intelligence of LLMs, making them far more capable than simple scripted IVRs.

Text-based bots were an intermediary stage between IVR systems (which used a voice interface but had limited capabilities) and the new generation of voice bots, which combine the intelligence of chatbots with a human-friendly voice interface.

In fact, voice bots are becoming so advanced that they remove a need in the Real-Time AI Assist technology in certain use cases.

With Real-Time AI Assist, the AI listens to conversations and provides real-time guidance to agents. For example, if a customer mentions wanting to cancel a subscription, the AI might prompt the agent to ask why and offer a discount. However, constantly bombarding agents with next steps, coaching suggestions, and sound bites can make it hard for them to concentrate on what the customer is saying. In some cases, this disrupts the natural flow of conversation and reduces agents to "voice bots," merely reading what the AI suggests.

If AI is already telling agents what to say and do, why not let AI speak directly to customers?

Don’t get me wrong—Agent Assist tools are incredibly powerful and invaluable for enabling agents to perform their jobs better and faster. But if the goal is to replace appropriate training programs, there may be a better solution: AI Voice Agents, or voice bots.

The mass adoption of AI Voice Agents is poised to significantly transform the industry.??

At MiaRec, we provide AI-powered evaluation of interactions between customers and agents. With more interactions being handled by voice bots, we’re seeing a shift toward evaluating customer-company interactions, regardless of whether they involve a human agent or a bot.

These are exciting times. Some compare this era to the dot-com boom.??

Now, we’re in the midst of an AI boom—and this time, we might just avoid the “bubble” part.