I’ve avoided “weighing in” on the subject of coding with AI, but after seeing people either dismiss AI tooling entirely or throw themselves at it without any structure, it felt appropriate to create a course to help. There’s a useful, fairly disciplined way to bring agents into a WordPress workflow, and I wanted to detail what that actually looks like in practice, rather than leaving people to figure it out through trial and error (or worse, “vibes”).
At Automattic we are lucky to have access to some of the best WordPress developers in the world and understand how agentic development” can be incorporated into a development workflow to save time and improve quality. Personally, the part I like the most is that an AI Agent doesn’t tire in the same way a human would, and so, with appropriate prompting, is actually less likely to miss details that a human might. At the same time, human review and accountability is also a vital part of the loop.
If that sounds like your kind of rabbit hole, the course is live now: AI WordPress Developer Tools and Workflow.
What’s in it
The course is in three parts.
The first part is just getting everyone on the same page – what an LLM actually is, where you’ll bump into one in your day-to-day work, what you should think twice about before sending real data off to a hosted API, and (importantly) when you genuinely don’t need an LLM at all and a plain script will do the job better.
The second part is probably the most useful for a typical working developer. It’s hands-on with Claude Code – the terminal agent, the desktop app, MCP for hooking Claude up to other systems, the Claude Code extension in your IDE, even the Chrome extension for working directly in WP Admin. There’s also a bit on Skills, and on building your own MCP servers. Rather than a bunch of disconnected demos, it all follows one example: a local WordPress site and a plugin that grows lesson by lesson.
The third part zooms out from “you and an agent on your laptop” to “your team’s actual repo and CI pipeline.” It covers planning AI-assisted work before any code gets written, a GitHub issue-to-pull-request workflow, what makes an issue something an agent can actually act on usefully, and how to keep the whole thing reviewable instead of just trusting the agent blindly. This part borrows heavily from a workflow originally built at Pew Research Center and opened up for anyone to use, which I think is a genuinely good pattern worth knowing about even outside of WordPress.
As with all courses on VIP Learn, it’s free, though sign-up is required.