Guava Raises $12.5M to Make Enterprise Voice Agents Reliable Enough for Real Calls

Voice AI developer platform Guava raises $12.5M led by Acrew to deliver reliable, compliant conversational agents for enterprise workflows.

The Acrew-backed company combines its own voice models with developer tooling for regulated customer conversations.

Guava has raised $12.5 million in seed funding led by Acrew Capital, with participation from Flourish Ventures, WillowTree Ventures, Clocktower Ventures and Commerce Ventures.

The funding was announced November 3. Guava develops APIs and software infrastructure for voice agents that can handle scheduling, customer support and other phone-based tasks.

The Los Angeles-based company is targeting a critical problem for enterprise adoption: a voice agent may sound convincing in a demonstration while failing to complete real customer requests accurately and consistently.

Building the Full Voice Stack

Guava offers a developer-first platform rather than asking customers to assemble speech recognition, language models, voice synthesis and telephony from separate vendors.

The company says it trains and serves its own models, allowing it to optimize latency and reliability across the entire interaction. Developers can integrate Guava's software into existing systems and control an agent's business rules through code.

Its products are aimed particularly at regulated sectors such as healthcare, finance and insurance, where conversations may involve sensitive information, compliance requirements and complex processes.

Guava reports more than ten billion live agent minutes across its technology lineage, over 100 regulated deployments and security attestations including SOC 2 Type II, HITRUST i1 and PCI DSS Level 1. Those are company-provided figures and compliance representations.

A Team From Marqeta and Gridspace

Guava was co-founded by CEO Omri Dahan, Anthony Scodary, Nico Benitez and Mike Piechna. Dahan previously spent more than a decade at Marqeta, where he served as chief revenue officer. The technical founders previously developed voice AI systems at Gridspace.

The company launched the Daytona voice model and a Voice Index benchmark to compare agent performance on tasks, including whether a call reaches its intended outcome.

Production Is the Test

Guava plans to use the seed capital for model development, product improvements and a larger go-to-market operation serving enterprises and business process outsourcers.

Its market opportunity is the gap between voice AI demonstrations and systems that can manage production calls day after day. But the company competes with model vendors, contact center platforms and fast-moving voice agent startups.

The most important measure will be reliable task completion at an attractive cost, not simply whether its agents sound human.

Sources: Guava / Business Wire announcement · Guava website

By Editor ·

18,600+ subscribers