Extreme Networks previewed Extreme Exchange zones, an on-premises enterprise AI architecture for branch offices, retail locations, campuses and other distributed sites. Planned for general availability in the first half of 2027, Exchange zones will use an Extreme Universal Compute Platform or qualifying customer-owned server alongside spare compute capacity in access points and switches. The company also reported an estimated 1,000 troubleshooting hours saved by customers during the first month of Extreme Agent ONE Coworker availability.

Exchange zones are designed to place AI applications and inference workloads according to enterprise policies governing latency, data residency and resilience. Each zone is bound to one site in one country. Models run locally by default, with cloud inference permitted only when policy allows and exceptions logged. The architecture targets operational telemetry and applications in networking, security, logistics and distributed operations, including scenarios where WAN connectivity is degraded.

Extreme has not disclosed hardware eligibility details, supported model sizes, inference benchmarks or resource isolation mechanisms. The promised economics of repurposing idle networking compute remain to be validated in production deployments.

Extreme separately reported that 41% of Platform ONE customers used Agent ONE for daily tasks in its first month, 98% of critical AI nudges received a response, and 58% of support cases initiated through Coworker included sufficient evidence to begin investigation. Engagement was roughly ten times that of the company’s earlier conversational AI agent. These are company-reported metrics; underlying sample sizes and measurement methods were not provided.

Agent ONE Coworker uses live network context and operational knowledge to surface proactive recommendations. Extreme cited Hill College, where staff used it to identify wireless access-point and firmware issues and restore service in about 20 minutes. The planned Agent ONE Operator mode, expected in H1 2027, would execute scheduled tasks and respond to events autonomously within defined governance boundaries.

“Exchange zones will put that compute to work, letting AI run locally on infrastructure customers already own,” said Nabil Bukhari, CTO and President of AI Platforms at Extreme Networks.

Analysis: Extreme’s Two-Year AI Networking Transformation

Extreme’s strategy has evolved from cloud-managed switching and wireless infrastructure toward unified network operations and agentic AI. The company introduced the AP5020 Wi-Fi 7 access point and 4000 Series switches in January 2024, then announced Extreme Platform ONE in December 2024 to integrate management, security, analytics and AI. Platform ONE expanded through 2025, with Agent ONE and its proactive Coworker capability emerging in 2026. This product sequence gives Extreme a common software and telemetry foundation across its enterprise networking portfolio.

Financially, Extreme reported approximately $1.14 billion in fiscal 2025 revenue and $1.284 billion in fiscal 2026, a 12.6% increase. The new Exchange zones concept extends the platform from AI-assisted management to local execution of AI applications, while Operator targets increasingly autonomous NetOps. Key questions for 2027 include the compute and memory available in deployed network hardware, workload scheduling and isolation, model compatibility, deployment costs, and whether measured operational savings translate into sustained software revenue. The announced adoption and productivity statistics are early company estimates rather than independently validated outcomes.

We’re tracking developments in enterprise networking and AI infrastructure. Follow our enterprise coverage.