AI agents·October 19, 2026, 21:31

Cisco wants AI agents to be teammates in Webex meetings

AI-generated from the sources listed, with automated source checks.

At WebexOne, Cisco showed a new Webex in which AI agents can be invited into meetings and spaces as part of the team. Several of the features, including Claude in Webex, are planned and have no exact date.

AI-generated image

On October 7, 2026, at the WebexOne conference in Austin, Texas, Cisco presented a new version of the Webex meeting software. The idea is that AI agents can be invited into spaces, meetings and calls on par with human colleagues.

An AI agent is a program that does not just answer questions but can carry out a series of tasks on its own, one after another.

Multistep tasks

According to Cisco, the agents should be able to handle multistep tasks across Webex and other applications. Cisco's own example is an analytics agent that pulls usage statistics and a presentation agent that then updates a PowerPoint with the figures.

Cisco also plans to give teams access to Claude in Webex. It will be built with Claude Managed Agents and should be able to analyze data, create content and do work using the company's shared knowledge, without users leaving Webex. The feature is listed as coming "soon," but no exact date has been given.

Jeetu Patel, Chief Product Officer at Cisco, says the next chapter of work is not about putting chatbots on yesterday's tools. It is about agents that work side by side with people and are embedded in their workspaces.

New for meeting rooms

RoomOS 27, the software on Cisco's meeting devices, is expected to be available in November 2026. Among other things, it is set to support Google Meet natively and include a multilingual AI assistant covering 32 languages.

Cisco also announced Dialog, an agent-based customer relationship tool, as well as Intelligent Workplace Experiences and the Cisco Ceiling Speaker Pro. Pricing for the new features is not stated in the sources.

A secure place for agents

When agents are allowed to act on their own, where they run becomes important. Perplexity has released its SPACE platform (Sandboxed Platform for Agentic Code Execution) as open source under the Apache 2.0 license. It runs agents in isolated virtual machines, meaning walled-off environments where an agent should not be able to reach the rest of the system. CoreWeave, Akamai and Crusoe are set to offer it to their customers.

According to Perplexity itself, none of nine leading models escaped from the virtual machine to the host in the tests that were run. Some testers did find ways around the network restrictions, however. The sources disagree on the number of attempts: 118 according to press coverage and 108 according to Perplexity's announcement. The tests are Perplexity's own.

Sources

More on this topic

Get the week's AI news in your inbox

Choose your level, topics and length. One email a week, unsubscribe at any time.

Subscribe to Promptly Newsletter