Coval raises $28M as enterprises pivot toward autonomous voice agents

The funding signals a shift in the AI stack from generic models to specialized evaluation infrastructure required for 'agentic' workflows. For streaming and communication platforms, reliable voice agents represent a massive efficiency play in customer service and technical support. As voice becomes a primary interface, the ability to simulate millions of interactions before deployment mitigates the brand risk associated with autonomous AI failures. Watch for whether Coval’s simulation-first approach, modeled after self-driving car testing at Waymo, becomes the standard verification protocol for real-time conversational interfaces.

The demand for robust voice testing mirrors a broader enterprise transition. Per Deloitte in February 2026, 25% of enterprises using generative AI have already deployed autonomous agents, a figure expected to double by late 2027. Gartner researchers noted in 2025 that conversational AI could handle over 50% of contact center volumes by 2027, potentially saving the industry $80 billion in labor costs during 2026 alone. However, deployment remains hampered by quality concerns, with one-third of professionals identifying accuracy and tone consistency as primary barriers to scaling.

Strategic shifts at major vendors reinforce this trend. Zoom launched its AI Companion 3.0 in March 2026, transitioning the tool from a meeting assistant to a 'system of action' capable of triggering autonomous workflows across Salesforce and ServiceNow. Similarly, Deepgram has integrated deeper observability tools into its voice infrastructure to counter the high failure rates of early-stage voice agents. Analysts at Gartner (June 2026) report that worldwide AI spending is on track to hit $2.59 trillion this year, yet enterprises are increasingly prioritizing 'governance architects' and simulation platforms over raw model innovation to ensure these investments yield measurable ROI.