AI coding tools have transformed how we work. Not as a gimmick, but as a genuine productivity multiplier for infrastructure and DevOps tasks.
Here's how we actually use them.
The Reality of AI Assistance
Let's be clear: AI tools don't replace expertise. They accelerate experts.
If you don't understand Terraform, Claude won't write good Terraform for you. But if you do understand it, Claude can help you write it 5x faster.
The sweet spot is experienced engineers using AI to eliminate tedious work and focus on the stuff that matters.
Infrastructure as Code
This is where AI shines brightest. Writing Terraform, Bicep, or ARM templates is often formulaic but time-consuming.
Example: Creating a Standard Azure Landing Zone
Instead of writing 500 lines of Terraform from memory (and looking up every resource attribute), we describe what we want:
"Create Terraform for an Azure landing zone with:
- Hub-spoke network topology
- Azure Firewall in the hub
- Bastion for management access
- Log Analytics workspace
- Diagnostic settings for all resources"
Claude generates a starting point in seconds. We review, adjust, and refine. Total time: 30 minutes instead of 3 hours.
FinOps Automation
We've built several tools for cost analysis using AI-assisted development:
Azure Cost Export Analyser - A Python script that:
- Reads Azure cost export CSVs
- Identifies anomalies and trends
- Generates recommendations
- Outputs a summary report
The core logic took us an hour to build with Claude's help. Doing it from scratch would have been a full day.
PowerShell and Bash Scripts
Quick scripts for ad-hoc tasks are perfect for AI:
"Write a PowerShell script that:
- Gets all Azure VMs across all subscriptions
- Checks if they have the Environment tag
- Outputs untagged VMs to CSV"
30 seconds of prompting instead of 15 minutes of writing and debugging.
Documentation
Nobody likes writing documentation. AI makes it bearable:
- Generate README files from code
- Create runbooks from scripts
- Write architecture decision records
- Produce handover documentation
We still review and edit, but the first draft appears instantly.
Code Review and Troubleshooting
"This Terraform is throwing error X. Here's the code and the error message. What's wrong?"
AI excels at pattern matching against known issues. Often it spots the problem immediately - a missing provider, a typo in a resource attribute, a misconfigured backend.
What Doesn't Work Well
Complex architectural decisions - AI can suggest options but can't understand your business context.
Novel problems - If the issue isn't similar to something in training data, AI struggles.
Security-sensitive work - We don't put credentials or proprietary information into AI tools.
Production changes - AI-generated code always needs human review before deployment.
Our Workflow
- Use AI for first drafts - Terraform modules, scripts, documentation
- Review everything - AI makes mistakes. Expert review is mandatory.
- Test thoroughly - AI-generated code goes through the same testing as human code
- Iterate with AI - "This doesn't handle case X, update it to..."
- Maintain human ownership - We understand and can support everything we deliver
The Client Benefit
Using AI tools means:
- Faster delivery on projects
- More competitive pricing
- Consistent quality (AI doesn't have off days)
- Better documentation (because it's less painful to produce)
We're transparent about using AI - it's a tool, like any other. The value we provide is expertise, judgement, and accountability. AI helps us deliver that value more efficiently.
Interested in how we can help with your Azure or DevOps challenges? Get in touch to discuss your project.