Kindred Voice

Kindred Voice develops agentic voice AI for enterprise telephony, built on native speech-to-speech models rather than a cascaded speech-to-text, LLM, text-to-speech pipeline. The company argues that conventional pipelines strip tone, emotion and accent from caller audio, add several seconds of latency, break in noisy environments, lack native multilingual support and misfire on tool calls beyond roughly six steps. Kindred's stated approach keeps audio end-to-end while placing a deterministic workflow layer above the model.

The product turns a customer's standard operating procedures and sample calls into graph-based workflows that are edited through a drag-and-drop interface. Kindred states that agents can be scoped and deployed to production in as little as 43 minutes, that non-technical staff can push changes and hotfixes in real time without redeploys, and that agents execute 20+ sequential tool calls per call including CRM lookups, transfers and follow-ups. Guardrails are described as enforced in code rather than in the prompt, with read-back of sensitive details, restrictions on unauthorized actions, phased rollouts, component testing, call logging for audit and connections to existing phone systems and CRMs. The system is presented as model-agnostic and natively multilingual across 100+ languages.

Reported deployments span property management, legal services and regulatory compliance: an AI receptionist across Anequim's 1,300-person property-management call center, after-hours client intake for a national immigration law firm, and a Mandarin-speaking regulatory survey agent built in about an hour by a non-technical operator through the UI.

Business model

Kindred Voice sells voice AI agents to enterprises on an outcome-based commercial model, handling workflow mapping, integration, voice tuning and compliance guardrails, and then allowing customer staff to maintain agents themselves. A self-serve tier was announced alongside the enterprise motion.

Outcome-based pricing charged per resolved call rather than per minute, at $1-$5 per resolved call depending on task complexity, with nothing charged for unresolved calls. A planned self-serve product offers a free tier with paid plans from $100 per month.

Traction

The company describes agents running in production across multiple industries, including a 1,300-person property-management call center where per-agent call volume tripled, a national immigration law firm handling all after-hours intake and expanding firm-wide, and a multilingual compliance agent built by a non-engineer in about an hour. A first signed customer is cited as running roughly 1,000 calls per month at about a third of prior cost.

Latest developments

A public teaser allowing users to build an agent by voice was live, with a full self-serve public beta stated to arrive in October, offering a free tier and plans from $100 per month.