Why Testing Infrastructure Must Become AI-Native

A New Chapter for Network Validation

Network testing has always depended on expertise. Whether validating a new data center fabric, benchmarking network equipment, or verifying application performance and security, engineers have traditionally needed deep knowledge of test tools, APIs, and product-specific workflows. As networks become larger and more complex, especially in AI-driven environments, that expertise remains valuable, but the operational model is starting to show its limits.

The challenge isn’t a lack of testing capabilities. It’s the growing gap between the speed at which infrastructure evolves and the number of specialists available to configure, execute, and analyze increasingly sophisticated tests.

At the same time, a different technology shift is underway. AI assistants, copilots, and agent-based workflows are becoming part of everyday operations across the networking vendor product development lifecycle and enterprise and carrier networking, and security teams. The question many organizations are beginning to ask is simple: if AI can help operate and troubleshoot infrastructure, why can’t it help run the tests that validate it?

Moving Beyond Traditional Test Workflows

Historically, professional testing platforms have required users to translate objectives into detailed configurations, scripts, and API calls. Those workflows provide powerful control, but they can also create barriers for teams that don’t have extensive product expertise.

As organizations develop and deploy larger AI data centers, higher-speed networks, and increasingly distributed applications, validation needs are expanding faster than specialist resources. What engineers often want is straightforward:

• “Run an RFC 2544 throughput test on the 400G ports using frame sizes from 64 to 9000 bytes and show me the binary search results for each frame size.”
• “Ramp traffic from 10% to 100% line rate in 10% increments and plot latency and jitter at each load level.”
• “Run a link flap test that brings the primary uplink down for 200ms every 30 seconds and measure recovery time after each event.”
• “Run the same latency test twice, once with UET and once with traditional Ethernet transport, then summarize the difference in a table.”

The intent is simple, but the path to execution is often not so.

A Different Approach: Conversational Testing

The emergence of the Model Context Protocol (MCP) is creating new possibilities for how infrastructure tools interact with AI systems. Rather than requiring users to work directly through GUIs, scripts, or APIs, MCP enables software capabilities to be exposed to AI agents as structured tools that can be discovered, orchestrated, and executed through natural-language interactions.

This concept is at the center of the VIAVI MCP Server Framework for Custom AI Workflows, which integrates TestCenter, CyberFlood, and TeraVM into AI-driven operational environments. Part of VIAVI’s NITRO® AI portfolio of solutions, the framework allows AI agents to configure, execute, monitor, and analyze test scenarios by translating user intent into actionable testing workflows.

The result is a shift from tool-centric testing toward outcome-centric testing. Instead of focusing on how to configure a test, engineers can focus on what they want to benchmark.

Why This Matters for Modern Infrastructure

The need for faster validation extends well beyond a single test domain. Network teams are being asked to validate increasingly complex AI fabrics and high-speed Ethernet environments. Security teams must continuously verify resilience against evolving threats. Application teams need confidence that services will deliver the user experience as expected under real-world conditions.

These activities often involve multiple tools, multiple teams, and multiple handoffs. By exposing testing functions as standardized AI-accessible tools, organizations can integrate validation directly into broader operational processes such as:

• AI-assisted network operations
• Security operations workflows
• CI/CD pipelines
• Infrastructure deployment processes
• Automated validation and assurance programs

Rather than existing as separate activities, testing becomes a more natural part of operational decision-making.

AI That Adapts

One of the more interesting aspects of AI-driven testing is the potential for adaptive workflows.

The MCP Server Framework supports real-time event streaming, enabling AI agents to receive live feedback as tests run. This allows workflows to evolve based on test outcomes rather than waiting until execution is complete. The framework also incorporates validation feedback mechanisms that help detect configuration issues before tests proceed, reducing wasted cycles and improving efficiency.

Importantly, automation does not mean removing people from the process. Human review checkpoints can be included at critical stages, ensuring that engineers maintain oversight where accuracy and accountability matter most.

Industry Recognition Reflects a Larger Trend

The industry’s growing interest in AI-enabled testing was evident at Interop Tokyo 2026, where the VIAVI MCP Server Framework for Custom AI Workflows received the Best of Show Grand Prize in the Network Infrastructure/Security/Testing category. The award recognized innovations that help organizations validate and secure increasingly complex AI, enterprise, cloud, and network infrastructure deployments.

While awards are noteworthy, the broader significance is what they represent: testing is beginning to evolve from a specialized activity performed by a limited group of experts into a capability that can be more broadly accessed throughout an organization.

Looking Ahead

AI will not replace network, security, or test engineers. Their expertise remains essential. What is changing is how that expertise is applied.

As AI agents become more capable participants in operational workflows, the ability to interact with professional testing systems through natural language and standardized interfaces will become increasingly important. Open approaches such as MCP help ensure that testing infrastructure can participate in those workflows without requiring custom integrations or vendor-specific automation.

To learn more about how AI-native test automation is helping organizations simplify validation workflows and integrate testing into broader AI-driven operations, download the MCP Server Framework for Custom AI Workflows brief.