For two years, enterprise AI has lived mostly in chat windows that sit next to your work. Cisco’s bet, announced at WebexOne 2026 in Austin on October 7, is that the real action is inside the app your team already has open all day. The company is turning Webex into a workspace where AI agents from both Anthropic and OpenAI can join meetings and chat spaces as working participants — not just to take notes after the fact, but to do tasks while the call is still running.
The headline move: Claude Managed Agents, Anthropic’s cloud-hosted agents, are coming to Webex. They can be added to group conversations as something closer to a team member than a bolted-on feature, taking on multi-step tasks in the background while the meeting continues.
What was announced
Cisco’s own example makes the pitch concrete. Someone @mentions an analytics agent in a Webex space to pull usage numbers, then asks a presentation agent to drop those figures straight into a PowerPoint deck — live, without anyone touching the file. The agents are meant to schedule meetings, track tasks and generate content using shared enterprise context already sitting inside the platform, according to Cisco’s announcement.
This is a step beyond where workplace AI has spent the last two years. Chatbots that summarize a transcript after a call are useful, but they’re reactive. What Cisco is describing is an agent that acts as a standing participant in the workflow — something you can pull into a conversation and hand a task to, the way you’d hand it to a colleague.
Cisco isn’t betting on one lab
Here’s the part that matters beyond Webex. Cisco isn’t making this a Claude-only product. Webex will also bring in OpenAI’s Dots agents, the always-on AI assistants OpenAI introduced at its DevDay event in September. So the two biggest names in frontier AI will sit inside the same collaboration app, available to the same enterprise customers.
The logic is easy to see from the buyer’s side. No IT department wants its collaboration suite locked to a single vendor’s model, and nobody wants to rip out their meeting platform the day their preferred AI lab pulls ahead. For Anthropic, landing inside Webex is a genuine distribution win outside its own chat interface. For Cisco, hedging across both labs is the obvious move.
Cisco is also building its own native agents into the platform. Personal Agents for Webex Calling can screen calls, converse with callers to understand their needs, determine urgency and take action when the user can’t pick up — they’ll arrive in the first quarter of 2027 alongside the redesigned Webex app. A Translator agent, previously announced, is now in controlled availability ahead of a general release in November, with real-time translation in ten languages including Hindi, English, French, German, Italian, Japanese, Korean, Mandarin, Portuguese and Spanish. And an AI Coach for Vidcast — Cisco’s short-form async video tool — watches a presenter deliver, reacts to pacing, tone and word choice as it happens, and asks the questions a live audience would. That one is available now.
Dialog, RoomOS 27 and the rest of the package
The agent layer doesn’t arrive alone. Cisco also announced Dialog, a new agentic harness aimed at customer experience rather than internal collaboration. It’s paired with a control plane called AI Agent 360, built to keep context persistent across a customer issue — so an agent keeps working a problem across channels and over time instead of resetting with every new ticket or call. Dialog is expected in beta in early 2027.
Then there’s the hardware side. RoomOS 27, the operating system for Cisco’s meeting-room devices, ships in November 2026. Its headline feature isn’t AI at all: native Google Meet support through a new partnership with Google, putting Meet alongside Webex, Microsoft Teams and Zoom on the same hardware. It’s an honest acknowledgment that most conference rooms now have to serve whatever video platform a given customer happens to be running. Cisco frames the three releases — the Webex agents, RoomOS 27 and Dialog — as covering how people work, how the physical workplace adapts and how businesses engage customers.
The refreshed Webex app itself gets a personalized landing page highlighting work priorities and upcoming meetings, suggesting follow-up actions like responding to messages or reviewing meeting recaps.
The reality check
Most of this is still preview. Cisco says the bulk of the new features — including the deeper Claude Managed Agents capabilities, the redesigned Webex app, collaborative agents and third-party agents — are planned for early 2027, with RoomOS 27 (November) the first piece to actually ship. The announcement provides no pricing and no paid-adoption figures; the architecture is real, but the product most customers will actually touch is still months out.
And there’s a harder question nobody at the event could answer yet: governance. Once agents can read company data and take actions across systems on an employee’s behalf, someone has to decide what they’re allowed to touch. Cisco has pointed to its Control Hub for observability, model selection and policy controls, and to AI Defense protection and Splunk monitoring as part of the package it can sell to enterprises — which is also where the company’s own upside lies. Cisco’s shares traded at $116.34 on October 9, up about 1.26% on the news, a modest move that reads as interest rather than conviction.
Still, the direction of travel is worth noting. The assumption for the past year has been that agentic AI distribution would be decided inside chat interfaces — ChatGPT, Claude.ai, Gemini. Cisco is betting it’s decided in the software people already have open all day. Whether the agents live up to the demo is an open question, but the battleground just got a lot more interesting.
Read more on the enterprise agent race:
GitHub Copilot Studio Turns AI Coding Agents Into Click-Built Workflows — the same move from transcripts to action, now aimed at developers.
Microsoft Decision-1: A 9B Decision Model That Claims 35x Faster AI Judgments — another sign the industry is shifting from chatty models to models that score and decide.
Instinct Raises $1B at $10B Valuation for a Personal AI Agent That Actually Does Things — the personal-agent bet, funded like the stakes are existential.
