There’s a gap between a governance document and colleagues genuinely knowing what’s expected of them. Reading twelve pages of policy isn’t training. It’s just reading.

I’ve done enough mandatory online training courses over the years to know what bad looks like. Walls of text, a multiple choice question you can answer without reading anything, a certificate at the end that means nothing. I didn’t want to build that. I wanted something short, clear, and actually useful. Colleagues finishing it and knowing the dos and don’ts of AI at the council. That’s it.

I built it using Claude Code, which made the whole thing significantly faster than it would have been otherwise. HTML first, then converted to SCORM so it could sit inside Rugby’s existing LMS rather than live as a standalone file somewhere.

What I couldn’t do alone was the rollout. HR handled deployment, identified who needed to complete it, and managed the comms. That side is easy to underestimate when you’re focused on building. It made a real difference to whether it actually landed.

It’s not perfect. But people complete it, they know what’s expected of them, and there’s a record that they’ve done so. For something built by someone with no background in training design, that feels like the right place to land.

But over time I kept getting pulled toward the stuff behind the interface. Why is this data so inconsistent? Where does this actually come from? Could we automate the bit where someone manually copies this into a spreadsheet every Monday morning? The questions that live one layer deeper than the UI.

The transition into data and automation engineering wasn’t a clean pivot. It was more of a slow drift that eventually became the whole job. And honestly, the front-end background has been more useful than I expected. Understanding how data gets presented shapes how you think about how it should be structured. Caring about user experience makes you a better designer of internal tools. The instincts transfer.

What’s changed the most in the last couple of years, though, is AI. Specifically how it’s changed what I can actually build on my own.

I’m not a classically trained data engineer. I don’t have a computer science degree. What I have is about a decade of getting things working in a browser and enough curiosity to have picked up Python, APIs, and automation tooling along the way. For a long time, that meant there were certain things I’d approach cautiously. More complex data pipelines, custom integrations, anything that required me to move quickly in a language or framework I wasn’t completely fluent in.

AI has changed that in a pretty fundamental way. I now use it as a genuine part of how I build things. Not as a shortcut, but as something closer to a collaborator. I’ll work through a problem in conversation before writing a line of code. I’ll use it to get a working skeleton up quickly so I can react to something real rather than plan something theoretical. And increasingly I’m building things that use AI APIs directly, tools that can process unstructured data, make sense of freeform text, and surface patterns that would have taken weeks to find manually.

The role I’m in now, leading a data and automation team in local government, would have looked pretty different without any of that. The scope of what’s achievable has genuinely shifted. Not because AI does the work, but because it closes the gap between having an idea and having something you can test, show, and build on.

I don’t think that replaces experience or judgement. If anything it makes both more important, because you can move faster, which means the quality of your decisions matters more. But it has made the work feel more open than it used to. Which, after a decade of mostly staring at CSS, is a good feeling.