TLDRocket
Sign in

Extreme Networks’ Agent ONE Coworker moves AI networking from dashboards to answers

SiliconANGLE Zeus Kerravala

Extreme’s new AI coworker is now live for Platform ONE customers. It watches for network trouble and pushes answers instead of making engineers hunt for them.

Based on reporting by SiliconANGLE, Zeus Kerravala — read the original for the full story.

Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error

Extreme Networks is trying to fix a very old habit in network operations: spot an alert, open a dashboard, chase logs, repeat. Its new Agent ONE Coworker, now generally available to all Platform ONE customers worldwide, is meant to shorten that loop by surfacing answers before an engineer goes looking for them.

The company says it learned the hard way. Nabil Bukhari, Extreme’s chief technology officer and president of AI, said the first wave of networking AI was mostly a chatbot sitting on messy, unconsolidated data. It looked good in demos. It fell apart in production. Extreme had 18 months of its own release data from the first version of Platform ONE to work with, and that fed into this launch.

The biggest change is under the hood. Extreme built a custom context layer and a knowledge graph that maps relationships across users, devices, applications, services and network conditions. That matters because the same label can mean different things in different parts of the product. Bukhari’s example was client ID, which appears in fabric, switching, Wi-Fi and software-defined wide-area networking, but doesn’t mean the same thing everywhere.

Agent ONE Coworker also behaves differently from the usual chatbot. It is ambient, always running, and uses a command-bar nudge to interrupt only when it has already done some initial investigation. Early results inside Extreme point to about a 900% jump in interaction in the first eight days. The company also built guardrails around that behavior: low-severity nudges can be snoozed or dismissed, but severity-one nudges cannot be snoozed.

There is another layer here: honesty. The agent can say when it cannot help, explain what it can do instead, and even package evidence and open a support ticket. That may sound basic, but in enterprise operations it is a big deal. A helpful hallucination is still a hallucination. Extreme’s bet is that engineers will trust a system more if it knows its limits and behaves like a teammate instead of a talkative tourist.

The rollout is deliberately cautious. Coworker mode keeps a human in the loop, while Operator mode, the autonomous tier, is due early next year. Extreme is also leaning on standard operating procedures and blueprints rather than pretending every network is a blank slate. That fits the company’s position: it is a challenger, not a giant, and it has to win on speed, not brute force. Fiscal 2026 revenue hit $1.28 billion, Platform ONE drove nearly half of subscription bookings, and the number of customers spending more than $1 million a year climbed to 187.

My take — AI-written commentary, not fact-checked reporting

This is the right way to ship enterprise AI: start as a nudge machine, not a remote-control gremlin. The industry keeps acting like autonomy is the prize, when the real prize is fewer minutes lost to digging through screens and logs. Extreme seems to understand that trust in ops is earned one boring, correct answer at a time.

Read more about this at: SiliconANGLE

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.