All the biggest headlines at Connect were about Extreme Platform ONE, a newly developed, integrated, cloud-based management platform that has been architected from the ground up to leverage agentic AI components and principles. First envisioned over two years ago, Extreme announced that Extreme Platform ONE had completed early availability and was ready for limited availability, with general availability expected in early calendar Q3 2025. As proof of this advanced stage of readiness, the Extreme team ran full live demos of Extreme Platform ONE throughout the event, with essentially no problems experienced.
The drivers behind the development of Extreme Platform ONE were many, including:
Staying at the forefront of AI innovation on behalf of Extreme’s customer and partners.
Building on and superseding existing AI components (such as ExtremeCloud IQ CoPilot) with a new solution that has the power to advance toward practical and effective automation, and even autonomy, enabled by exposed model reasoning and a human-in-the-loop approach.
Providing consistent levels of functionality and support across the entire Extreme networking product portfolio with a single, seamless monitoring and management solution. More specifically, by eliminating variations and different look and feel between switch, fabric, and Wi-Fi management tools.
Laying the groundwork for future expansion and innovation, to participate in workflows that go well beyond just the network.
Further delivering on the promise of powerful simplicity for enterprise networking.
The vision and design behind Extreme Platform ONE sought to overcome a number of challenges with existing AI-driven solutions in the marketplace. These include fractured tooling, limited capabilities (i.e. just natural language queries), “add-ons” that sit on top of existing tools but are not fully integrated, and build-it-yourself API-centric approaches that create development load and can result in brittle outcomes. Extreme’s solution is aligned to cover three important pillars:
1. Strong data infrastructure. In this case, that means data from all Extreme product categories, including specific local context of each individual customer’s networks, plus the best-practices knowledge base from Extreme’s design and support services groups.
2. Appropriate agency controls. Here, the focus is two part—first to ensure appropriate guardrails are in place for AI functions, and second to achieve that by having agents inherit the authorizations (same or any lesser access) held by whomever invokes the agent. This simplifies governance by adhering to existing human-oriented policies. Full audit logging here also helps to ensure transparency and traceability.
3. Effective orchestration capabilities. The approach here is to assign a lead agent for any task or inquiry but allow multiple agents to interact along the workflow path to completion. This simplifies communications with the human in the loop, while the execution process discloses which sub-agents are engaged and the results of their activities.
The result is a well-executed management platform that offers intuitive data presentation, easy navigation, and AI assistance embedded throughout. Feedback from early availability users, both customers and partners, was positive. Some of the key Extreme Platform ONE capabilities detailed at the event included:
Service Agent for most direct operator interactions. This starts with full natural language chat but also includes suggestions of corrective workflows and the ability to automatically execute them as approved.
Agent Manager for configuring various agents and for adding/building new agents and data connectors.
AI Canvas, for AI-assisted building and sharing of custom dashboards. This is a big step forward from the fixed dashboards that are part of the current ExtremeCloud IQ management solution and drew positive feedback from conference attendees.
Support for monitoring non-Extreme/third-party network devices.
New multilayer topology mapping and visualizations, which for the first time bring together physical, logical, access, RF (radio frequency), fabric, and service views across Extreme-based networks, even including third-party devices, with correlation across layers and side-by-side comparison.
Embedded security analytics to help leverage the network behavior viewpoint for supporting incident detection and forensics.
Integrated support for managing inventory and licensing across the deployed device fleet.
The scope of capabilities within Extreme Platform ONE fits well with the findings of recent Enterprise Strategy Group research findings regarding the objectives and values of AI and automation for networking.1 For instance, top use cases for AI in networking include network capacity planning, network performance optimization, anomaly detection, and predictive maintenance, all of which can be addressed by Extreme Platform ONE. Further, those organizations who have successfully tied AI and network automation together, as is made possible with Extreme Platform ONE, reported smoother updates, improved stability/uptime and better overall security.
1. Source: Enterprise Strategy Group Complete Survey Results, The Role of AIOps in Network Infrastructure Operations, April 2024.