On this page
TRIOTECA is a mortgage AI infrastructure company for the Spanish market and one of Spain’s leading digital mortgage brokers. Its technology powers AI agents that automate and improve the mortgage journey at scale.
Over the last few years, Trioteca has evolved from a digital mortgage broker into a mortgage AI infrastructure company. That transformation has shaped how it builds, scales, and chooses partners. Today, Trioteca runs seven AI agents around the clock. One of them, Maria, places around 1,000 calls a day to prospective borrowers. She handles a workload that previously required a large team, allowing human specialists to focus on the moments where expertise creates the most value. Over the past year Trioteca doubled its output with the same team and the same marketing budget, demonstrating how AI can create scale without proportional increases in headcount.
The Challenge
When Trioteca started building Maria three years ago, the voice AI landscape looked very different. Latency was the first problem. Speech models were slower, text-to-speech sounded synthetic, and the round trip from a caller's words to the agent's reply left pauses that broke the flow of a real conversation.
Ricard Garriga, Trioteca's founder, built the first version in-house and tested what else was available. Results were inconsistent: some providers had acceptable latency but poor voice quality, others sounded natural but could not handle calls at scale. For an agent that needed to hold many conversations at once, that gap between demo-grade and production-grade was the whole problem. Over three years the core challenge shifted from latency to concurrency. Trioteca needed one platform that could carry the load, and a partner it could count on when something went wrong.
The Solution
Trioteca runs its production voice stack on Telnyx. SIP Trunking connects its cloud to the Spanish PSTN with local numbers and carrier-grade termination. The Voice API handles call control, routing inbound and outbound calls and managing recording for compliance. Telnyx Voice AI Agents run the conversations, and Telnyx Inference serves the language models that understand mortgage inquiries and generate replies in real time. Telephony, AI, and inference sit on one platform, with one support channel behind them.
That choice was sealed not by a sales process but by how Telnyx handled a hard moment. The day before Christmas, Trioteca hit a critical outage and its agents stopped taking calls. With the technical team in a panic and a holiday weekend closing in, Garriga sent a LinkedIn message to David Casem, Telnyx's CEO. Casem replied in under a minute. It was not an automated response: it was the CEO himself, on the eve of a holiday, pulling in his team. The two teams worked together through Christmas Day until Trioteca was running again. "I had the feeling that your team was my team," Garriga says, "because you were so worried about our problems."
The word he returns to when asked why Trioteca chose Telnyx is TRUST.
The Results
- Revenue doubled year over year, with the same team and the same marketing spend.
- The AI agents grew from about 36% of Trioteca's Telnyx usage in June 2025 to roughly 77% in June 2026.
- Capacity scaled without headcount. Maria places around 1,000 outbound calls a day, while the other agents qualify leads, verify documents, answer borrower questions, and escalate complex cases to staff.
- Telephony, AI, and inference run on one platform with one support channel, so problems are easier to isolate.
Trioteca is not slowing down. It is building more agents and evaluating Telnyx Ultra voices as a Spanish-language TTS option, testing whether the newer models can match or beat its current voice quality. That evaluation runs through the same shared Slack channel where Trioteca's engineers work with Telnyx support, a workflow that started during the Christmas incident and has not stopped since.
For Garriga, the bet on agents is part of a larger view about where the edge actually is. Technology, he argues, is becoming a commodity: the models and the infrastructure keep getting better and are available to everyone, so they are no longer where a company wins. What is left is focus and partners.
Trioteca's ambition is simple: become the world's leading mortgage AI company. To achieve that goal, it focuses relentlessly on mortgage expertise while relying on best-in-class partners for the underlying infrastructure. "You need the right partners to grow, with the same mentality," he says. "Choosing the right partners is one of the most important things you can do as a founder."
His operating philosophy is simple: try first, learn fast. "You have two options: wait or try. We are on the second. Maybe we're wrong, maybe we make mistakes, but we learn. And for other companies thinking about whether to do it or not, they are late."
The technology will keep improving and arriving for everyone, but what Trioteca found in Telnyx is the thing that does not commoditize: a partner it can trust to keep the infrastructure on the frontier, so the team can stay focused on being the best in the world at mortgage AI.
