Coldi Deploys Multilingual AI Agents for Sales Calls

What changed

Israeli startup Coldi is deploying multilingual AI voice agents for outbound and inbound sales calls. Founded in 2025 by Or Goldshmidt and Ilia Ronin, the company has raised $500,000, employs 15 people and says Hebrew entered production in September 2026 with its first Israeli customers.

Its system combines voice agents with telephony, CRM and calendar integrations, campaign management, quality monitoring and live transfers to human sales teams. Coldi targets regulated, high-volume callers including fintech, insurance, brokerage and debt-collection businesses, charging per minute plus a support fee after a paid pilot.

Why it matters

Coldi is betting that the valuable part of a sales call is not necessarily the first conversation. Its agents can qualify leads, re-engage prospects, call people who abandoned onboarding, book meetings and pass promising calls to humans.

That shifts where salespeople enter the process. Humans remain responsible for the conversations most likely to need judgment, while the software handles the repetitive calling around them. The practical test is whether the handoff produces useful opportunities, not whether the AI sounds human.

The company says its agents work in more than 30 languages and that it has clients across Europe, the Middle East, Latin America and South Africa. Hebrew is its newest production language, giving its Israeli launch a concrete milestone rather than just a localization promise.

Coldi’s larger claim is architectural. A voice model alone is not enough for a business that must connect calls to records, calendars, phone systems and staff. Its proprietary orchestration layer, Campaign Brain, is meant to run that whole chain.

The outlook is promising but still unproven: Coldi has disclosed neither customer counts, call volumes, revenue nor conversion rates, and there is no reported evidence yet that its agents outperform human callers or competing systems. For now, the company has shown a managed product, a $500,000 funding base and an expanding language footprint. The harder proof will be what happens after the phone is answered.