The voice channel in the contact centre has lost love in recent years. However, recent developments in voice AI (generative AI, speech recognition and speech synthesis) might be finally bringing voice into the 21st century. Voice AI is now turning the traditional one-to-one, costly contact channel into a one-to-many, scalable, revenue-generating one.
In a recent discussion with Sam Rosendorff, VP, Global Presales, boost.ai (recently crowned Gartner Magic Quadrant for Conversational AI Platforms Leader), we discussed the resurgence of voice AI, why the time is now for businesses to implement it, and the strategic benefits of doing so.
What’s changed in voice AI?
There have been a couple of pivotal developments in voice AI over the last few years, even in the last few months, that have set the foundations for voice AI finally being able to deliver on its promise.
- Generative AI. The general advancements in generative AI has meant that callers can be understood with far greater levels of accuracy. Rarely will you ever hear those dreaded words “I’m sorry, I didn’t understand that”.
- Retrieval Augmented Generation (RAG). Being able to search through masses of content with high levels of accuracy means that you’re able to answer a much broader range of customer queries.
- Automatic Speech Recognition. The ability to successfully transcribe what the caller says is crucial in sending the right information to your model. Recent advancements in Speech to Text have seen Word Error Rates (WER) as low as 1% through distillation and fine tuning.
- Speech Synthesis. Synthetic voices have come on leaps and bounds even just in 2025. No longer does your voice AI sound robotic, and no longer do you need to pre-record voice actors. You can clone voices with ease or select from thousands of off the shelf, high quality options.
According to Sam, “You can’t just put a wrapper around a chatbot and make it into good voice AI, and you can’t just use an LLM and make good voice AI, because hallucinations are a problem.”
When you consider that many organisations had already successfully deployed and scaled NLU-based voice AI, adding these developments into the mix means we now have a toolkit that, arguably, isn’t lacking in any area. Natural, human-like voice AI experiences are now entirely feasible for large enterprises.
Why is voice a prime channel for AI automation?
One of the biggest challenges with AI adoption is exactly that… Adoption. Deploying a chatbot on your website doesn’t guarantee anyone will use it. Most chatbots sit quietly in the bottom corner, yearning for clicks. Gaining adoption of your chatbot relies on your user taking the initiative and deciding to engage. Of course, there’s lots you can do to increase usage with things like proactive pop-ups and such, but you’re still always on the back foot, trying all you can to start a conversation.
Voice doesn’t have that problem.
The moment a customer picks up the phone and dials your contact centre, you have their full attention. As soon as your voice AI agent answers the call, you have adoption. It’s baked in. It’s the default.
As Sam explained, “It’s true, but it’s also the way that we’ve trained people for decades to speak. They just want to be able to get an answer or help with what they need. It’s no longer about what we think they want; it’s about giving them the experience they’ve dialled in for.”
Of course, the caller may simply say “I’d like to speak to an agent”, which we’ve observed can happen in up to 40% of cases. However, even in the worst case scenario, you have 60% of all calls automatically engaging with your voice AI agent.
Turning a “I’d like to speak to an agent” into increased AI usage
Those that object and want to speak to an agent immediately, they’re not a lost cause. Because they’re already talking to your voice AI agent, you have an opportunity to keep the conversation going. You have a chance, at the very least, to understand why they’re calling by responding with something like:
“Sure, I can transfer you to one of my colleagues. To make sure I put you through to the right person, can you please tell me why you’re calling?”
At this point, if the user answers your question, you can figure out whether you can help them. For example, let’s say your user says:
“I need to speak to someone about the bill you’ve just sent me.”
If your voice AI agent can help with billing queries, you can respond with:
“No problem. I can actually help with billing queries and save you having to wait to speak to a colleague. If you could tell me more about your specific question, I’ll see if I can help.”
At this point, the user is either going to play the game and tell you more, which means you’ve successfully continued the conversation and increased the chance of a successful automated conversation. Or, the user will object again and ask to speak to someone.
You don’t need to automate everything from end-to-end to see ROI
You also don’t need to automate everything from end to end to get value. I’ve explained previously the various ways in which voice AI can contribute to customer outcomes and business success without needed to conduct end-to-end automation. Granted, that’s where you’ll probably end up, but that’s not where you need to start to find value.
Take our above example. Let’s say the user objects to having a conversation and still wants to speak to a human agent. Here, all is not lost because you now know what the user needs. You know their intent.
That means you can route the caller to the most appropriate team and send a summary of the conversation so far to the agent that answers the call. You’ve just gathered intel on contact drivers, reduced average handle time and removed any potential for internal call transfers.
This is a profound but often overlooked advantage of voice AI. The voice channel is in the critical path of your customer service operation. It’s where urgency, intent and emotion converge and where even small efficiencies can create disproportionate value.
Beyond cost reduction to revenue
That value, however, isn’t always confined to cost saving. In fact, most AI deployments, if executed well, will have both a cost-to-serve impact, as well as a growth impact.
One of the clearest examples of this in practice is IKEA. In several markets, IKEA has deployed voice AI at the front door of its contact centre. Billy handles nearly 50% of incoming calls successfully.
Did IKEA sack its staff and cash-in the savings? Nope. It retrained its staff to become interior design consultants and now charges customers for their time. It turned a cost centre into a revenue centre.
Yes, voice AI can deflect (or as I prefer to say; triage). Yes, it can automate (or, as I prefer; serve). But it can also scale your brand, your service standards and your customer promise, without scaling your headcount.
As Sam put it: “We’re seeing a shift where companies aren’t just chasing cost savings. They’re looking at how AI can help them stand out by truly understanding their customers and giving them great experiences.”
Don’t wait for perfect: start where you are
So where does that leave you? Well, if there’s one thing that holds most organisations back from deploying voice AI, it’s the belief that they’re not ready.
But you already have the customers. You already have the call volume. You probably already know the top reasons people get in touch. And you likely have the workflows, APIs and data to handle a good portion of those queries with automation.
You’ve also, highly likely, got a lot of room for improvement: long wait times, high AHT and so on. How much longer are you going to leave yourself in the state that you’re in?
You don’t need a perfect use case or a full transformation strategy to start. You need a focused entry point, a capable partner, the right technology and a clear commitment.
To dive deeper into more practicalities and learn how to deploy voice AI effectively, including design best practices, how to deal with the dreaded silence of latency, why you can’t just slap a voice in front of a chatbot, and much more, check out the full discussion between myself and Sam at boost.ai.
