The operational burden of managing enterprise campus and mobility networks is becoming unsustainable. Siloed tools, disparate operating systems, and a lack of real time visibility leave network administrators trapped in a reactive loop of fighting fires instead of managing systems. Arista Networks is tackling this operational challenge by extending its unified software philosophy into the campus fabric and anchoring it with AVA, an agentic AIOps platform. By pairing a single binary operating system approach with autonomous troubleshooting capabilities, Arista appears to be shifting the paradigm of network operations from manual investigation to guided, intelligent remediation. We had an amazing opportunity to learn how AVA works recently during Mobility Field Day.
AVA-ry Interesting Methodology
At the core of Arista’s campus strategy is its long standing commitment to a unified software architecture. In practice, running the same Extensible Operating System (EOS) binary across data center switches, campus hardware, and wireless access points ensures absolute data fidelity. This architectural consistency eliminates the telemetric gaps that traditionally plague multi vendor or multi operating system environments. Taken together, these clean data streams feed directly into Arista’s Network Data Lake (NetDL), creating a high fidelity observability fabric that serves as the single source of truth for the entire network state.
This unified data foundation is what allows AVA to function as an autonomous agent rather than a passive chatbot. When a user logs a connectivity issue, a webhook from a ticketing system like ServiceNow can alert AVA to open an autonomous session immediately. Driven by an internal state machine, AVA issues API calls across the Arista suite, querying Arista Guardian for Network Identity (AGNI) for network access control logs and CloudVision for endpoint history before a human engineer ever looks at the ticket. This background investigation ensures that when an operator opens the trouble ticket, all necessary diagnostic context and log summaries are already compiled and attached.
To navigate complex troubleshooting scenarios, AVA relies on a structured repository of human authored guides called skills documents. For example, AVA utilizes an SSID profile skill to analyze wireless configurations and a troubleshoot client skill to investigate local radio frequency issues. If an older corporate asset suddenly fails to connect, AVA can cross reference these skills to diagnose the failure. The SSID profile skill might reveal that legacy 802.11b connection rates were recently disabled, while the troubleshoot client skill pulls the access point logs showing a non supported rates error. By correlating the data between these two distinct skills, AVA identifies the exact root cause with high confidence.
Beyond configuration checks, the troubleshoot client skill is leveraged to rule out wireless issues, proving the fault lies downstream. AVA can pull authentication statuses, device segmentation profiles, and switch ACLs to map the entire path from the client to the application. If a user cannot access a repository, AVA might reveal that the local Wi-Fi path is perfectly clean but a specific firewall rule is intentionally blocking access to GitHub. This multi domain visibility quickly shifts the focus away from the wireless infrastructure, accelerating the true time to resolution.
Radio Resources To the Rescue
Proactive validation represents another significant layer of this architecture, driven by hardware innovations in Arista’s Wi-Fi 7 access points. These units feature a dedicated, tri-band fourth radio, a multi-function radio designed to temporarily convert itself into a simulated Wi-Fi client. This radio connects over the air to a neighboring access point, pushes security certificates to authenticate via EAP-TLS, grabs an IP address, and performs synthetic client testing against critical applications. AVA can initiate these synthetic tests as background tasks to validate network health right before a business day begins, or to verify that a newly implemented firewall change restored hop by hop application performance.
Despite this deep autonomy, Arista enforces a strict human in the loop safeguard for all remediation actions. AVA is engineered so that it cannot push configuration modifications or execute change control scripts independently. Instead, it packages the proposed fix, such as a script to modify a load balancing profile or remove a blocking ACL, and explicitly prompts the operator for approval. This interaction model respects strict Role-Based Access Control (RBAC), meaning a user with read only permissions cannot leverage AVA to alter the network. If an engineer rejects the recommendation, the change control is abandoned, leaving the production network untouched.
Bringing IT All Together
Arista’s implementation of agentic AI emphasizes a fundamental truth in modern enterprise networking: automation is only as good as the underlying data fabric. By combining a single binary OS approach with NetDL, Arista avoids the fragmented telemetry that causes other AI tools to hallucinate or misdiagnose issues. AVA does not represent a replacement for network engineers, but an acceleration mechanism that handles the administrative tax of log scraping and context gathering. For enterprise organizations, the value is clear. Shifting from reactive firefighting to an autonomous, human approved troubleshooting workflow drastically lowers the mean time to resolution and allows senior operations teams to focus on strategy rather than tracking down legacy rate mismatches.
To learn more about Arista and AVA, make sure you check out their website at https://Arista.com. To see their entire presentation from Mobility Field Day, head over to their presentation appearance page here.
