Zoom launched Zoom Virtual Agent 3.0 (ZVA) today, building directly on the AI Companion 3.0 architecture released in December 2025.
Where AI Companion brought agentic workflows to meetings, documents, and daily tasks, ZVA 3.0 extends the same execution engine to multi-step, cross-system workflows — completing business actions end to end rather than handing off to a human at each decision point.
Gartner predicts that 40% of enterprise applications will include task-specific AI agents by the end of 2026, up from less than 5% in 2025. By 2029, it expects at least half of all knowledge workers to need new skills to work with, govern, or build AI agents. For IT and UC leaders, the question is no longer whether to adopt agentic AI — it is which platforms can deliver it without creating new governance headaches.
“Agentic AI was just the beginning,” said Chris Morrissey, General Manager of Zoom CX.
“Zoom Virtual Agent 3.0 orchestrates multi-step workflows across systems, continuously learns from human resolutions, and provides full transparency into every agentic action. This allows organizations to confidently automate complex interactions.”
Why Most Enterprise AI Still Doesn’t Execute
There is a gap between what enterprise AI promises and what it currently does. Most tools deployed across Zoom Workplace, Microsoft 365, and similar platforms act as assistants: they draft, summarise, suggest, and retrieve. A human still has to make decisions and take action. Agentic AI closes that gap by letting the system carry a task through from start to finish, across multiple back-end systems, without waiting for human input at each step.
Zoom Virtual Agent 3.0 is Zoom’s attempt to do this at enterprise scale. Its enhanced AI execution framework runs multi-step workflows across compatible CRM, billing, and order management systems in a single uninterrupted flow, eliminating the hand-offs and context switches that typically slow complex processes down. Zoom has already tested this internally: its no-match rate dropped from 35% to 0%, and self-service resolution on its billing team rose from 0% to 30% in three months, saving over 1,000 staff hours per month.
The Governance Feature That Could Matter Most to IT Teams
The second feature available now is agent journey transparency and governance. Account admins can see the data sources, decision logic, and workflow paths behind every action the agent takes. Teams can audit outcomes, trace failures, and refine automation rules without having to pause deployments or escalate to vendors.
For IT leaders, this directly addresses the most common reason agentic AI stalls in enterprise rollouts: when automation acts on real business data across live systems, organisations need to know exactly what it did and why. The World Economic Forum has pointed to transparency as a baseline requirement for scaling AI responsibly, and increasingly enterprise procurement teams agree. Observability is not an advanced feature, it’s the condition under which scaling becomes possible.
Three Capabilities Arriving in Spring 2026
Zoom has confirmed three further features for general availability in Spring 2026:
- Multimodal LLM intelligence will let ZVA read and act on documents, images, and structured inputs such as serial numbers: handling tasks that currently require a human to manually extract information before automation can proceed.
- Continuous learning will build a feedback loop from resolved human escalations, applying those outcomes to similar future requests with oversight controls in place.
- Proactive outbound engagement will allow the agent to act on known events without waiting for a user to make contact, reducing inbound volume before it builds.
What This Means for Teams Running on Zoom
The most practical angle for organisations already standardised on Zoom Workplace is that ZVA 3.0 runs on the same model layer and governance framework as AI Companion 3.0. There is no separate implementation, and no additional AI stack to manage. Enterprises get agentic execution across both internal workflows and external-facing processes through infrastructure they already operate.
That shared foundation is what separates this from a standalone product launch. As Microsoft Copilot, Google, and others compete for the same ground, Zoom’s bet is that consistency across the platform is more valuable to enterprise IT than individual feature wins.
“It’s more than a product update,” said Morrissey. “It’s another step toward more connected relationships, where AI and humans work together to resolve issues faster and build trust.”
