Cisco AgenticOps

IT admins looking after meeting rooms and collaboration devices often have to jump between several tools and datasets to resolve a single issue. Across a global fleet, with small IT teams and no one onsite, that makes triage slow and difficult.

Cisco Workspaces MCP helps admins ask the AI agents they already use about their rooms and devices, in plain language and in real time. They can check workspace and device health, look up event history and error codes, compare configurations, see room conditions such as temperature and noise, and review how workspaces are used.

Queries run against live Control Hub data, so the agent always reasons over the current state of the rooms.

• Microsoft Copilot Studio connector
• Gemini Enterprise connector
• Cisco AI Assistant in Control Hub (Webex blog)

As a designer, I worked in a team of software engineers and a project manager.

• Led the experience design for how AI agents work with Cisco Workspaces data
• Researched with IT admins how they triage room and device issues
• Defined AI principles for trustworthy agent behavior, covering transparency, uncertainty and human-in-the-loop, implemented in the Workspaces MCP
• Contributed to the agent behavior of the Cisco AI Assistant in Control Hub
• Contributed to the AI strategy for Control Hub with product and design partners, bringing device and workspace use cases, prototypes and feedback into the discussions

• Worked with engineers on defining the tool architecture: which tools exist, their scopes and descriptions, and how the agent combines several tools
• Designed error states, feedback loops and elicitation flows for when data is missing, unclear or uncertain
• Designed generative UI, widgets and MCP app concepts
• Tested and iterated on tool descriptions together with engineering
• Built a repeatable agentic behavior evaluation method through simulations
• Prototyped with realistic device and workspace data in code, using Claude Code, Codex and Cursor