Enterprise Voice AI That Actually Works

Pressing zero to reach a human should not be the default plan. We talk with Fred Fontes CEO from Acclaim about what’s finally making enterprise voice-first AI work in the real world, especially inside regulated industries like banking and financial services where compliance, auditability, and data security are not negotiable. For teams burned by old IVR trees and brittle chatbots, the conversation gets practical fast: what has changed in the underlying models, and what has to change in how we deploy and control them.

We dig into the idea of sovereignty and why many CIOs and CTOs feel trapped between the need to innovate and the risk of sending sensitive customer data through multiple third-party clouds. Fred explains how controllable voice AI agents, strong guardrails, and enterprise-grade orchestration can turn “cool demos” into dependable contact center automation. We also get into domain-specific benchmarking, because a universal speech-to-text score does not matter if you cannot accurately transcribe a noisy telephony call about banking topics.

Then we go beneath the hood on outcomes: banking collections use cases showing six to eight percentage points higher recovery rates, the ability to A/B test messaging quickly, and why interaction costs can drop dramatically when conversations are faster, more accurate, and handled in parallel. We also talk about the human side, shifting agents toward higher-value customer experience work, and the hardest obstacle left: integration with systems of record and enterprise workflows.

If you’re building or buying conversational AI, listen closely, share this with a teammate who owns CX or security, and subscribe, leave a review, and tell us what your biggest blocker is to deploying voice AI at scale.

Transcript

Welcome And Guest Introduction

SPEAKER_01 0:01

Hey everybody. Fascinating and timely topic today as we dive into the world of enterprise voice first AI with a real innovator in the space at Acclaim. Fred, how are you? Doing well, Ellen.

SPEAKER_00 0:14

Good morning.

Fred’s Path From Consulting To Voice AI

SPEAKER_01 0:14

Thanks for having me. Well, thanks for being here. You spent many years, decades maybe in CX and AI. Perhaps before we dive into Acclaim, introduce yourself and your journey. And how do you describe Acclaim?

SPEAKER_00 0:28

Yeah, absolutely. As you said, I've been in the AI space, applied AI space for well over a decade at this point. Started in the world of consulting at Deloitte and McKinsey, helping Fortune 500 companies deploy AI in all parts of their organization, trying to find value and data back when AI wasn't called AI and ML was the cool world word of the day. And then spent the last seven years in the startup world, first in the world of supply chain AI, helping organizations move goods around the world faster and more efficiently at a company called ERA in the decision intelligence space. And in the last three years, truly immersed in the world of CX at Replicant Building Voice AI agents and joined the claim about two months ago to launch the brand in the US and globally and start taking this product to market.

Why Voice-First AI Works Now

SPEAKER_01 1:22

Fantastic. Really exciting time, of course. We're all immersed in these topics. But what's changed recently, in your opinion? Recently could be a couple of years or so, that makes voice first AI really viable at scale in the enterprise and really compelling.

SPEAKER_00 1:39

Yeah, absolutely. It's funny to talk in years these days. It feels like everything is changing week over week. Um, but what we really believe is change and that enables this technology is a combination of the underlying models having gotten truly, truly quite phenomenal, um, as well as the technology that brings all of these models together to deliver the quality that both the consumer expects, being able to get answers to their questions, being able to get their problems resolved, not having to repeat themselves and try to figure out what the AI is expecting to hear. Um but just as importantly for the enterprises, AI that is controllable, that is manageable, that is not going to disobey guardrails, um, and especially in our space, uh AI that treats compliance, treats data security as a first-class citizen as it needs to be.

Why Legacy Automation Missed The Last Mile

SPEAKER_01 2:33

Fantastic. And you know, automation has been around for, gosh, decades, uh, RPA for the last decade and other techniques. But why has legacy automation kind of failed to deliver on a lot of the promises made? And um what is really new and radical here with your approach?

SPEAKER_00 2:53

Yeah, absolutely. I I think the the challenge that we've seen with legacy automation is it it has never really connected the last mile. Uh, we've managed to automate chunks of the process and we've made those chunks incredibly efficient. But at some point, you always needed to effectively go to a human or go to human-to-human interaction uh to finally close the loop. And what we think is radically different today is really that ability to automate not just the business process, but the interaction required to really close the loop and get the action to be driven all the way down to the consumer or all the way down to the partner.

Sovereignty Compliance And Guardrails

SPEAKER_01 3:31

Yeah, fantastic answer. Um, and you you know a lot about the enterprise. What does it mean in practice? There are a lot of voice bots. I think I saw a market map of like 3,000 voice bots out there. But when you get into compliance and auditability and orchestration and control, security, I mean, a lot of these are non-negotiable. They're all non-negotiable for enterprise use. How does that impact your vision and plan?

SPEAKER_00 4:01

For us, it's it's a question of really making sure that those principles are at the foundation of what we build. And the way we think about this is at the core sovereignty. It's the ability to deploy all of these models soup to nuts uh within the four walls of an organization and help, you know, what I would call the CIOs, the CTOs, or the innovation leaders move out of this place of being between a rock and a hard place of I must innovate and I must introduce AI. But the only way today I have to do so is by sending my data to a third provider, to a third party that's in the cloud, who's then sending that data to their providers who are also in the cloud. And there's you know, many times uh dozens of them for any given company. Um, so really it starts with sovereignty, and then on top of that, sovereignty, uh, it's really that auditability, control, guardrails, making sure that when we say the model is going to do a certain thing, or when we say the agent uh is gonna follow a particular process, uh then we can guarantee that outcome.

Fixing IVR Baggage With Proof

SPEAKER_01 5:07

Great. And as you know, voice AI still has a bit of a reputation problem starting back in the uh IVR days. Sadly, too many of those are are still out there, plus you know, kind of the failed uh chatbot experience that many of us hate. Um we're continually trying to press zero to talk to a real person. You know, how do you convince organizations that this time, you know, something is fundamentally different?

SPEAKER_00 5:35

For us, it starts with hearing is believing, uh, our tagline that on our website is results that speak for themselves, and and we really we really stand behind that, um, is being able to experience the the automation, being able to experience how it's different. Um, for us, it's also a question of being able to demonstrate for our specific domain how do the results look look different. And what I mean by that is, you know, again, I mentioned earlier, we own our own stack, we build our own transcription models, we fine-tune our own LLMs, we build our own text-to-speech. In all of those areas, we're always working against benchmarks that are specific to the work we do. So when we say our transcription accuracy is higher than market average, uh, we mean it's higher than market average on a voice conversation that's running over telephony lines that is discussing banking topics because that's what matters uh to a bank, right? It's great to look at word error rate benchmarks that are universal and those are great. But uh in our world, you know, it doesn't matter if we're able to transcribe YouTube audio accurately. What matters is are we able to transcribe that conversation that a caller is having with their bank on banking topics with a high degree of accuracy?

Banking Collections Wins And Cost Impact

SPEAKER_01 6:53

Brilliant. And what are some of the big operational wins your customers are seeing right now? Use cases or verticals. Maybe you can share some anecdotes there.

SPEAKER_00 7:05

Yeah, absolutely. The I would say the the main topics uh that we're really seeing benefit in are uh both kind of better results or better outcomes. So we've got several clients in the banking space where we work uh with them on the collections uh vertical or collections topic, where they're seeing six to eight percentage points higher uh recovery rates than working with human agents. Uh part of that is again the controllability and the accuracy uh that we can put behind the agents and the models. It's also our ability to rapidly A-B test messaging that works and be much more agile in how we run the process. Um I think it's a given in the world of AI agents these days. Obviously, there's a massive cost benefit uh to AI agents over human human workforce. Um, so anywhere from five to 10x reduction in cost.

SPEAKER_01 8:00

Wow.

SPEAKER_00 8:01

Um, but I would say the the part that you know I really and the team really get energized around is the impact on the human workforce. Uh, in many cases, we're not seeing our clients uh you know reduce the size of their contact center or reduce the size of their workforce, but they're really taking that existing workforce and they're able to focus it towards higher value added tasks, more concierge-like interactions with customers, driving customer lifetime value, in what I would argue was the whole point of having a contact center, uh, not really how do I manage to deflect as many conversations as possible.

How 90 Percent Interaction Cost Drops

SPEAKER_01 8:38

Yeah, great point. You talk about reducing interaction cost by about 90%. That is amazing. Maybe give a peek underneath the hood. What is happening to drive that level of efficiency?

SPEAKER_00 8:51

Yeah, absolutely. The conversations are faster, they're more accurate, they're more to the point. Uh, we're able to have more of those conversations in parallel uh than teams of of human agents would. Um they're, as I mentioned earlier, higher accuracy and better outcomes. Uh, we're able to iterate on the conversation, the speech, the uh even the voices uh that we want to use to be able to drive those better outcomes. Um and then a big part of this too is again, we're using our own models. So we're not paying for third-party providers and having to pass along the cost of an ASR or an LOM or a PCS.

Keeping CX Human With AI

SPEAKER_01 9:32

Yeah, that's brilliant. So fine-tuning is clearly an opportunity. Uh and what what's your philosophy around the balancing act between all this incredible automation with customer experience? So interactions are really more human, not less human, with the use of AI. Mm-hmm.

SPEAKER_00 9:51

Yeah, I think it goes back to that comment I made earlier around how do we focus our human workforce to really do what they excel at. It's having those nuanced conversations, building trust and rapport with customers, uh, really solving challenges that AI is not equipped to solve. And in many cases, making judgment calls that AI shouldn't be making in the first place. Um, really enabling them to offer service, uh, not necessarily just followers for it. Brilliant.

Why Voice AI Is A Real Category

SPEAKER_01 10:22

And with your investor hat on, you know, what signals are you seeing that voice AI is a really serious enterprise category of its own? Um, it's been a long time coming, but there's been you know tons of funding announcements and news and private equity and mergers in particular uh happening in this space. How do you rationalize all of that and what's going on?

SPEAKER_00 10:43

Yeah, I mean, I think it's recognition that it's a problem that's worthwhile going after and uh the time is ripe to do so. Um, you know, when you look at global spend on CX, the world spends anywhere between$250 to$350 billion between tech and people on the problem. And I think uh, you know, I would uh argue both sides of the equation aren't happy with the outcome. Consumers hate having to pick up the phone and call 1-800 number, and on the other side, you've got an agent that is burned down and underpaid, uh, you know, and counting down the seconds until they can jump to the next conversation. Uh, so I think, again, it's recognition that this category is worth pursuing at the time is right. Um, I think what you're seeing though is also a little bit of consolidation from the days of, hey, it's now easier because I can kind of quote unquote solve the problem by stitching a few APIs together. Uh, and the recognition that from a technical standpoint, from a change management standpoint, uh, from a deployment standpoint, it's a harder problem than some people thought it would be.

Go To Market With Customers And Integrators

SPEAKER_01 11:50

Indeed. That's uh well said. Um maybe talk about your go-to-market for those who may not be familiar with you. Uh, how do you work with customers? How do you work with partners, integrators, and what does your sort of delivery model look like?

SPEAKER_00 12:07

Yeah, absolutely. So we're we're just launching, uh, as I mentioned, in the US, we just launched the brand yesterday uh as part of the process. Um, and we go to market in in both both ways, both direct, uh working directly with our customers to deploy uh these AI agents and solutions, um, either cloud-based or within their four walls, um, depending on their security posture and needs, um, as well as working with third-party integrators to deploy the technology. It really, you know, on our side, we're here to build the best technology we can, and we're here to build the AI agents that solve the problem, soup to nuts. Uh, but then how that gets integrated into our customers' uh environments, we're we're agnostic on whether customers want to do that themselves, get our help, or work with a third party.

Integration As The Real Bottleneck

SPEAKER_01 12:56

Fantastic. So time to market has always been a challenge. Time to revenue, let's say, with the voice AI, the amount of tweaking and integration required and reluctance to make changes to workflows have all been barriers. Um, how do you think about breaking down those barriers and silos and getting these services to market faster than traditionally we've been able to do?

SPEAKER_00 13:19

Yeah, it's it's the biggest problem, and honestly, at least from my experience, the biggest challenge that customers run into is building the AI agent is no longer the bottleneck. It's how do I connect it to the data and the systems that it needs to be effective. Uh, we're thinking about this in a couple of different ways. One is really building those end-to-end solutions, uh, working with partners within our ecosystem. We are very focused on banking, financial services, and other regulated industries. Uh, so being able to work with the handful of systems of record in those industries, the handful of uh CCAS's in those industries so that the integrations are effectively out of the box. Um, and secondly, really is thinking about how do we innovate uh on that integration topic? How do we use AI agents to solve that problem for us? Uh, whether that's topics like browser use, uh, if human agent has an interface that they can use, why can't an AI agent use that same interface? Um, but also how can we build integrations using AI agents faster than before?

US Launch Strategy And Conferences

SPEAKER_01 14:22

Brilliant. Well, that's uh gonna be greatly appreciated out there in the marketplace. Looking ahead over this year, what are you excited about in terms of your go-to-market? Uh, anything on your agenda or radar that uh you might be interested in sharing?

SPEAKER_00 14:36

Yeah, we're we're really excited again about the the launch in the US. We're building up our go-to-market team here quite quickly uh based in Miami, uh, taking a little bit of a different approach uh than I would say some of the folks in Silicon Valley, uh, both from the investment standpoint. Uh, we're funded by Ratmir Timeshev, the co-founder of Veeam, uh, who funded our$34 million series A, uh, but also taking a bit of a longer-term view to the market. Uh, it's not about raising as much capital as we can as quickly as we can and buying space in the market, but really building technology that, you know, as you mentioned, is easy to integrate, easy to deploy, delivers value quickly and delivers real long-term uh sustainable value. Um, and then looking forward to meeting folks at conferences. We're getting ready for a conference in Vegas here next week in terms of uh Credit Union Collections Association. Uh, we're also getting ready to show up at CCW and reveal the brand there for the RCS community.

SPEAKER_01 15:36

Fantastic. We'll look forward to seeing you and the team out in the marketplace and congratulations on all the success and more success to come. Thanks, Red.

Closing Thanks And Where To Follow

SPEAKER_00 15:45

Absolutely. Thank you, Evan.

SPEAKER_01 15:47

And thanks, everyone, for sharing this episode. Also check out the TV show, techimpact.tv on Bloomberg Television and Fox Business. Thanks, Red. Thanks, everyone.

SPEAKER_00 15:57

Thanks, Evan. Take care.