In a recent episode of the Tech on Deck webinar, Inbenta CTO Merlin Bise spoke with TSIA’s George Humphrey about the rapid evolution of voice AI, what’s actually working in production today, and the guardrails that enterprises must put in place to succeed. Voice AI in the enterprise refers to AI‑powered systems that understand and generate natural speech to automate customer and employee interactions across phone, apps, and devices. It's rapidly moving from pilots to mission‑critical infrastructure for support, sales, and operations.
The following is an edited transcript of their conversation.
Merlin Bise is Chief Technology Officer at Inbenta. He brings over 20 years of experience in technology leadership, with deep expertise in combating fraud through innovative solutions and designing a sophisticated OCR platform. His forward thinking and creativity have allowed him to build multiple AI platforms that serve some of the largest corporations in the world.Merlin formerly co-founded GIACT Systems (acquired by LSEG) and during his tenure, he drove new product development and implemented technology across some of the largest global enterprises, including Equifax, Walmart, Costco, PayPal, Intuit, AIG, Quicken, PNC Bank, American Red Cross and federal and state entities.
What is Voice AI in the enterprise?
It’s really a conversational solution that’s intelligent. You think about traditional phone systems, they’re very much rules-based.You’re going to get fixed answers. It’s controllable. But customers get frustrated because, unless you invested a lot of money or were using one of the really large IVR systems, the system had issues keeping up, it lost track of previous conversations, and you could go off the rails.
Voice AI has opened up a whole new world. It’s intelligent. It has the ability to think. You certainly can build compliance and controls around it, so it’s safe. but it can impact a variety of places in your business, including internal usage. It’s way beyond just external. So to me, it’s a branch of what you’re trying to accomplish with not only your customer success but customer service.
Is Voice AI going to have the same impact on IVRs and contact center solutions that generative AI is having on chat?
Yeah, absolutely it will. When large models came on the scene, it was flipped on its head. We had to pivot and build more intelligent systems. And I think the same thing with voice, you’re going to see a huge transition because you don’t have to worry about call trees. But you really have to do a shift to compliance. You’ve got to make sure you’ve got the right safety controls. You’ve got to be able to gracefully bring a human in if necessary during a conversation. And you’ve got to audit everything.
You’ve got to make sure you’ve got the right safety controls. You’ve got to be able to gracefully bring a human in if necessary during a conversation. And you’ve got to audit everything.
But what you can do with it and achieve… a customer could call you concerned with a support issue, you could address that. They could then pivot right in the middle of that conversation and say, “Oh, by the way, I want to upgrade my account.” Well, guess what? They can do it.
What are some real-world use cases for Voice AI in the enterprise?
There’s a restaurant chain, very large in Europe and the U.S., and one of the challenges they have is when the customer picks up their order and it’s not hot or they’re not satisfied. Right now they call a number, and the chain is handling phone calls for about nine bucks a click.
Let the customer talk to a voice solution, get a resolution, and you keep it under a dollar easily — and you’ll have all that data in a beautiful database.
So here was my strategy. Let’s incorporate voice AI: scan the receipt, capture all the details, let the customer talk to a voice solution, get a resolution, and you keep it under a dollar easily — and you’ll have all that data in a beautiful database. You’ll know if they try to commit fraud or use the same receipt again. We're saving them about 85% and it’s more accurate.
The other one is a large credit bureau. We’re making their Knowledge Center intelligent, then building in the capability of voice and digital access to that content, and refreshing it so it’s always accurate. The consumer can have a real conversation in real time, fully audited, with sentiment analysis and immediate escalation to a human if it feels like there’s any unresolved issue.
How should CTOs evaluate Voice AI vendors?
Number one, you’re going to have to stay compliant. Compliance has got to be critical to you. If you’re dealing with medical data, HIPAA is going to apply. As a standard, you should meet ISO certifications. You should be using models or third parties that actually have SOC and HIPAA capabilities and PCI.
Then beyond the normal regulation, where the data resides is critical. You’ve got to make sure you have safety controls. It’s one of the biggest mistakes companies make: Don’t trust AI to know when to say no — you’ve got to have controls there.
Then audit trails are critical.
And then always — always — have a human backup.
Related Articles
