Case study
AI-assisted internal development
From my work leading IT at a mid-sized manufacturer. Names are left out on purpose.
Context
A mid-sized manufacturer wanted IT to build and change its own tools faster, without putting its core business system at risk.
My role
I led IT and did most of the hands-on work myself.
Approach
I set up a version-controlled workflow in Git so every change is tracked and reversible. AI coding tools helped draft the code. Anything that touches the ERP uses read-only access patterns, so a tool can look at data but cannot change it. I also ran hands-on training sessions so colleagues in process improvement, accounting, compliance, cloud administration, contracts and the leadership team could use the tools themselves.
What I built
Production monitoring and reporting tools, a customer service order-checking tool, customer service call monitoring, a digital board for the paint line, and an API layer for a legacy ERP.
Tools
Git, Claude Code, Microsoft Copilot, SQL.
Result
More than 170 commits across 20 repositories in five weeks, mostly my own work. My own project cycles shortened from about a month to about a week.
That code was not peer reviewed, so this shows speed and workflow discipline, not an independently audited result.
Lessons
- Build a framework for managing projects and code from the start, so team members can collaborate on development and learn from each other's feedback. The whole team's knowledge improves, not just one person's.
- Set up an internal development team to handle internal projects. Departments are usually busy, and they benefit from a team that comes alongside them and supports the work.