Zoom Expands Agentic AI Workflows

🔑 Enhanced Key Takeaways

• •Zoom’s Federated AI architecture, a distributed reasoning system, powers these agentic capabilities and achieved 92.8% first-pass accuracy on retail-domain tasks in Tau²-bench evaluations[2].
• •New enterprise search includes 10 secure connectors for third-party platforms like Salesforce, ServiceNow, Box, Google Drive, and OneDrive to integrate external data with AI Companion[6].
• •Zoom developed a Small Language Model (SLM) with state-of-the-art performance in the 2 billion parameter category, customized via AI Studio for efficient agentic tasks[5].
• •AI Companion 3.0 features a dedicated tab in Zoom Workplace with a redesigned interface for consistent access across desktop, mobile, and web[6].

🛠️ Technical Deep Dive

• •Federated AI uses a distributed reasoning architecture combining small language models (SLMs) and larger models from OpenAI and Anthropic for orchestrated, cost-effective performance[2][5][7].
• •Custom SLMs (2B parameters) achieve state-of-the-art benchmarks, customized through Zoom AI Studio (low-code drag-and-drop builder) for specialized workloads like machine translation and agentic tasks[4][5].
• •Zoom Virtual Agent maintains conversation context across multiple questions using natural language understanding, with expansions to multilingual support and Zoom CX Insights analytics[1].
• •AI Expert Assist 3.0 and Customer Workflow Orchestration enable shared customer intelligence across service, sales, and marketing[2].

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline