Extreme Networks says customers using its Agent ONE Coworker assistant saved about 1,000 hours of troubleshooting in its first month. The company is also planning to turn spare capacity in installed network equipment into on-site AI infrastructure, a more ambitious bet than another chatbot for IT teams.
Watch Desk analysis
What happened
A SiliconANGLE guest column by analyst Zeus Kerravala describes two developments from Extreme’s AI Summit. The company says 41% of Platform ONE customers used Coworker for daily tasks during its first month, and users responded to 98% of the agent’s critical alerts, which Extreme calls “nudges”. Those are company-reported figures, not independently measured results in the account.
The larger plan is Extreme Exchange zones, previewed for the first half of 2027. They would combine a server, either Extreme’s or a customer’s, with spare computing capacity in installed access points and switches. The company says applications from its marketplace would run locally on open-weight models by default; cloud inference would be used only when policy permits, and logged each time. Extreme also plans an autonomous Operator mode for Agent ONE in the same period.
Why it matters
Network gear is usually there to move data and keep services running. Extreme’s proposal would make it part of the AI computing infrastructure too, using capacity already installed at a site. That could appeal to organisations concerned with where data is processed, how much telemetry they send to the cloud, or whether a service can keep working when connectivity is limited.
It is a proposal, not a broadly available service: the zones are planned for 2027, and the column says the demonstration applications were built with design partners. The reported Coworker figures offer an early usage signal, but do not establish how much time customers save across deployments or how the planned infrastructure will perform.
Our read
The useful idea is that the network might become both the thing AI manages and a place where some AI runs. That could give network teams a bigger role in local computing, provided the applications, controls and economics add up. For now, treat the usage numbers as Extreme’s early account and the Exchange zones as a promise with a date attached. In AI, 2027 is not tomorrow, despite what the calendar says.
What to watch
- Whether Exchange zones arrive in the first half of 2027, and which hardware they support.
- What applications customers can run locally, and what server capacity they need.
- How policies govern cloud inference and the promised logging.
- Whether Extreme publishes deployment results that substantiate its early Coworker figures.
Discussion spark: Should organisations use spare capacity in network equipment for local AI, or is it better to keep network infrastructure focused on connectivity?
Sources and evidence
- Extreme Networks wants the network to run AI, not just be run by it – SiliconANGLE (30 October 2026, 19:53 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.