Why We Chose to Build Our Own Voice Technology

Voice could make business AI easier to use, but third-party constraints around concurrency, flexibility, and unit economics pushed us to investigate building our own system.

Some customers would rather talk than type

Typing is not the easiest way to communicate for everyone. A customer may prefer to call, explain what they need, and get an answer naturally. For a business, voice can make support more accessible and create a more direct route to a useful conversation.

The economics had to work

As we explored voice for ChatMoni, we evaluated external voice platforms, including ElevenLabs and Cartesia, alongside other available options. These services offered useful capabilities, but we had to consider more than voice quality alone. Concurrency limits, usage costs, flexibility, operational control, and the economics of offering the feature to our own customers all mattered.

A dependency can become a product constraint

Relying on a third-party service can be a sensible way to move quickly. It can also create limits when the product depends on high-volume, real-time conversations. If demand grows, capacity rules and per-minute pricing can influence the experience and leave little room to build an affordable service on top.

The decision to investigate our own engine

By November 2025, we made the strategic decision to pursue our own voice technology. This was not the easiest route. Building speech systems requires work across audio generation, streaming, infrastructure, latency, quality, and deployment. But it offered a path toward greater control over how the service could scale and how its costs could evolve.

A difficult choice with a clear purpose

The objective was not to build technology for its own sake. It was to make voice a practical part of the ChatMoni experience, with economics and infrastructure that could support real businesses. That decision set up one of the most technically demanding chapters of the journey.