Building Safer and Smarter DevOps Workflows with Agentic AI

Introduction

Continuous learning has always been central to my career in IT. Having worked extensively with Microsoft 365 administration, identity and access management, endpoint management, Azure services, and cloud security technologies, I have learned that technical excellence comes from continually adapting to new tools, engineering practices, and emerging technologies.

Week 2 of the DevOps Training on Agentic AI provided an excellent opportunity to explore how artificial intelligence can become an active participant in modern software engineering and infrastructure automation rather than simply serving as a conversational assistant.

Claude Code as an Engineering Assistant

One of the most valuable experiences this week was working with Claude Code directly inside the development workflow. Rather than generating isolated code snippets, Claude Code became part of the engineering process by assisting with documentation, project navigation, troubleshooting, configuration, and infrastructure tasks.

This reinforced an important principle: AI delivers the greatest value when combined with engineering judgement, verification, and structured workflows. The technology accelerates delivery, but responsibility for design, validation, and decision-making remains firmly with the engineer.

Skills, Subagents and Modular Thinking

Another concept I found particularly valuable was the use of Skills and Subagents. Delegating specialised work to focused AI agents closely mirrors how experienced engineering teams organise expertise across infrastructure, security, automation, and operations.

Breaking larger problems into smaller reusable workflows encourages modular thinking, improves maintainability, and creates solutions that can scale alongside increasingly complex projects.

Working with Model Context Protocol (MCP)

Configuring Model Context Protocol (MCP) servers proved to be one of the most challenging aspects of the week. Since I was working within a managed Windows environment, I encountered differences in Docker configuration, package installation, Windows PATH configuration, and tooling compatibility compared to the standard course demonstrations.

Although resolving these issues required additional investigation, they reflected real-world enterprise environments where engineers frequently adapt solutions to organisational standards instead of relying solely on tutorial examples.

"The most valuable learning often happens during troubleshooting rather than during successful execution."

Hooks, Permissions and Safety Rails

Another highlight was implementing Hooks, Permissions, and Safety Rails within Claude Code. Configuring safeguards to prevent destructive commands demonstrated the importance of governance in AI-assisted engineering workflows.

As organisations increasingly integrate AI into DevOps and cloud engineering, balancing automation with security, operational control, and responsible governance will become an essential engineering practice.

Claude Memory

Exploring Claude Memory demonstrated how project-specific knowledge can persist across sessions. Retaining architectural decisions, project conventions, and implementation guidance improves consistency while reducing repetitive context-setting during long-term engineering projects.

Key Takeaways

• AI is most effective when integrated into disciplined engineering workflows.
• Engineering judgement remains more important than automation alone.
• Security and governance must accompany every automated workflow.
• Real-world environments require adaptability beyond tutorial examples.
• Documenting implementation and troubleshooting accelerates continuous learning.

Looking Ahead

As I continue developing my expertise in DevOps, Cloud Engineering, and Cloud Security, I see Agentic AI becoming an important complement to my existing Microsoft cloud experience. Combining Infrastructure as Code, automation, and AI-assisted engineering has significant potential to improve productivity, consistency, and operational quality when implemented with appropriate governance.

Week 2 has reinforced that successful AI adoption is not simply about using new tools. It is about integrating them into established engineering practices that prioritise security, maintainability, collaboration, and continuous improvement. Those principles will continue to guide my learning journey as I build expertise across Azure, DevOps, and cloud security.