Zoom has added new AI capabilities to its Virtual Agent customer service product, including a tool that generates voice and digital agents from simple text prompts and a suite for monitoring how those automated agents perform. The updates target support teams that want to deploy automation without manually building complex workflows and need hard data on whether the AI resolves customer issues.
Agent Architect turns prompts into agents
Agent Architect lets teams create automated agents by describing what the agent should do, rather than building step-by-step conversation flows. The tool interprets intent, fills in missing context, and connects to relevant systems and data sources to produce a customer service journey ready for review. Teams can refine the agent's behaviour before launch to match business requirements and customer expectations.
Zoom says these agents can handle more complex requests than traditional scripted bots. They gather additional information during a conversation, adapt to customer needs, and take actions across connected systems. This shift toward agents that manage multi-step tasks reflects broader momentum in AI Agents & Automation within customer service.
Monitoring agent performance
The Agent Performance Suite gives customer experience teams visibility into how well automated agents work in practice. It combines testing, quality management, and knowledge base support. One component, Agent Performance, simulates customer interactions before deployment and compares those results with live production outcomes. Dashboards show metrics including resolution rates, containment, and cost per resolution.
Quality Management for Zoom Virtual Agent applies the same assessment framework across AI-led, human-led, and mixed interactions. This lets organisations compare service quality across support channels and spot where customers get stuck. When connected with Zoom Contact Centre, KB Suggestions identifies successful human-assisted resolutions and drafts knowledge base articles for review, aiming to improve self-service content over time.
Unified context and local deployment
Zoom is adding a customer context layer that carries interaction history across Zoom Virtual Agent, Zoom Contact Centre, and Zoom AI Expert Assist. Customers no longer need to repeat information when moving between automated and live support. That context builds over time and can inform routing decisions, recommendations, and agent guidance.
Multi-location deployment lets organisations build one AI-driven service experience and roll it out across sites while localising phone numbers, greetings, routing, and knowledge bases. This is aimed at sectors with distributed operations - retail, healthcare, campuses, and franchise networks - where a central team maintains governance but individual locations tailor responses to local needs.
An optional outcome-based pricing model ties charges to resolved interactions or interactions successfully routed across voice and chat, rather than relying solely on conventional billing methods.
Chris Morrissey, General Manager of Zoom CX, described the thinking behind the update: "AI has significantly accelerated the CX landscape, and organisations not focused on outcomes fall behind. It's no longer just about deploying it to drive efficiency, but about having the context to drive personalisation at scale. But the challenge is eliminating the tradeoff between speed and sophistication, and Zoom CX bridges that gap so teams can personalise better, deliver faster, and drive stronger outcomes."
Why this matters for customer support teams
The new tools give support teams tighter control over AI for Customer Support, from agent creation to ongoing quality checks. Instead of trusting a black box, teams can test automated agents before launch, compare them against live results, and gradually improve self-service content based on what actually resolves customer issues. The shared context layer also reduces repetition for customers and agents, making handoffs between bots and humans less jarring.
Your membership also unlocks:
