Voice agents in minutes, compliance off the plate: Kindred Voice builds on Telnyx

Anyone who has tried to put an AI agent on a real phone line knows the trouble often starts before the first word is spoken: number registrations, carrier rules, and support tickets that nobody answers.

Kindred Voice set out to remove those obstacles for its own customers. "We built a self-serve, no-code platform where anyone can build voice AI for any definable workflow within minutes," said Phillip Hwang, CEO of Kindred Voice. Getting there meant choosing a telephony partner that could handle the parts of voice AI that never show up in a demo.

Choosing the provider that answered

Kindred Voice had no incumbent voice AI stack to replace. Telnyx was the foundation the platform was built on, and the path there ran through two other providers.

"We started out with OpenPhone, but they didn't have voice AI back then," Hwang said. "We tried going with Twilio, but Twilio's customer support was not that good, and we couldn't get the 10DLC certifications that we needed, and we couldn't talk to a human."

Compliance was the deciding factor. Business texting in the United States runs through 10DLC registration, and for a young company, an unresolved registration can stall everything behind it. "We chose Telnyx because it's the easiest entry with the compliance requirements," Hwang said. "The compliance requirements, like 10DLC, are easiest to resolve there."

The people mattered as much as the process. "Telnyx was very responsive," he said. "Josh, shout out to Josh, he helped us through the process." Once Kindred Voice was on board, the question of switching never came back up.

A reliability layer on top of the network

Telnyx carried Kindred Voice's first production voice AI launch. "We didn't have a voice AI solution before Telnyx," Hwang said. "Telnyx was our first one to actually launch this."

On top of the Telnyx network, Kindred Voice runs its own harness, a model reliability layer that keeps agents on task. That combination is what lets the company promise speed to its customers. "It only takes a few minutes now with us," Hwang said.

He also credited the building blocks Telnyx offers teams that want to work directly on the platform. "I know that Telnyx also is a very stellar voice AI platform," he said. "They give you all the model providers and all the infrastructure in one go."

Latency that kept getting shorter

For a voice agent, every extra fraction of a second on a call sounds like hesitation. Hwang has watched that gap narrow since Kindred Voice started on Telnyx in July of last year.

"Since then I've seen latency really shortening. I've seen reliability increasing," he said. "Telnyx has made massive leaps and bounds to become one of the dominant players in the space."

He attributed the change to where Telnyx has pointed its engineering. "I see that Telnyx is really focused on stabilizing their servers and using a lot of their engineering resources so that the intra-network latency is minimized," he said. "I've noticed a significant improvement just by using Telnyx over the past year and two months."

Servers close to the caller

Kindred Voice builds for callers across North America, and distance shows up on every call. "Building in North America means that you have to go closer to the customers," Hwang said.

That is where the network design pays off. With SIP trunking, Telnyx places its servers near the people on the line, and a single configuration choice handles the routing. "You can just pick a setting called latency, and it will automatically route to a server that's closest to the call, which minimizes latency," he said. "Telnyx is putting a lot of engineering effort into making that more reliable and lower latency."

Why Kindred Voice recommends Telnyx

For Hwang, the recommendation comes back to what happens when something breaks. Real-time calls are unforgiving, and eventually something goes wrong for every provider.

"I think the focus of Telnyx on really customer success is the biggest reason," he said. "Telephony connections are very latency sensitive, and a lot of things can go wrong, and they inevitably will. But when something does go wrong, you want a team that knows what happened, why it happened, and gives you a response."

He pointed to an incident in February, when Telnyx servers briefly went down after what he described as a missile strike on an AWS data center. "That is something that you can't really prepare against," Hwang said. "I was able to receive updates really quickly, and in turn reassure my customers, and get a really fast action plan from Telnyx. That kind of responsiveness and trust is what I need for a business that handles these calls."