Uploaded August 2026 | Updated September 2026, 2 weeks ago
Getting started with an agentic harness to troubleshoot your infrastructure is much simpler than you think. You don't need "Skills," and you don't need "MCP servers exposing everything as a tool." You can just start with any agentic runtime and add capabilities as they are needed. You also don't need to automate everything; in fact, “how to bring down your infrastructure with fully autonomous agents” is the opposite of what this talk is about. We want to show you how to leverage AI to decrease MTTR (mean time to resolution) and how to proactively monitor your infrastructure. That said, fully autonomous operations will be treated as a special case, and we will explain how you can begin identifying those use cases.
By the end of this session, attendees will understand:
• How to place any troubleshooting setup on the operator, harness, and autonomy map, and why a bare conversation is often enough to begin.
• When to add a tool, a skill, or persistent memory, and how to tell which gap you are actually facing before you over-build.
• Why autonomy is a property of the harness rather than the model, which is what makes the harness the place your risk and your safety budget live.
• What a realistic first step toward autonomous operation looks like, and the four gates an action class should clear before you let it run unattended.
Presented by Erik Smith, Dell; Rohan Mehta, Micron Technology
Learn More:
• SNIA Website: snia.org
• SNIA Educational Library: snia.org/library
• X: twitter.com/SNIA
• LinkedIn: linkedin.com/company/snia
Getting started with an agentic harness to troubleshoot your infrastructure is much simpler than you think. You don't need "Skills," and you don't need "MCP servers exposing everything as a tool." You can just start with any agentic runtime and add capabilities as they are needed. You also don't need to automate everything; in fact, “how to bring down your infrastructure with fully autonomous agents” is the opposite of what this talk is about. We want to show you how to leverage AI to decrease MTTR (mean time to resolution) and how to proactively monitor your infrastructure. That said, fully autonomous operations will be treated as a special case, and we will explain how you can begin identifying those use cases.
By the end of this session, attendees will understand:
• How to place any troubleshooting setup on the operator, harness, and autonomy map, and why a bare conversation is often enough to begin.
• When to add a tool, a skill, or persistent memory, and how to tell which gap you are actually facing before you over-build.
• Why autonomy is a property of the harness rather than the model, which is what makes the harness the place your risk and your safety budget live.
• What a realistic first step toward autonomous operation looks like, and the four gates an action class should clear before you let it run unattended.
Presented by Erik Smith, Dell; Rohan Mehta, Micron Technology
Learn More:
• SNIA Website: snia.org
• SNIA Educational Library: snia.org/library
• X: twitter.com/SNIA
• LinkedIn: linkedin.com/company/snia
