Work / Custom build
AI inside WordPress: job classification and generated answers
Two plugins that put language models to work inside WordPress: imported job ads sorted into the right categories, and visitor submissions answered under an editable prompt.
What they needed
Two clients wanted a language model doing a specific, narrow job inside their WordPress site, with the prompt in the editors’ hands rather than in the code.
What I built
Job classification (2025). A job board imports its vacancies from outside. When an imported job is saved, the plugin sends its text to an OpenAI model with a prompt the editors control. The model answers in JSON: discipline, education level, job type and weekly hours. The plugin creates any missing terms and files the job under them, so the filters work without anyone tagging by hand.
Generated answers (2024). Visitors submit a short text through a form. It goes to GPT-4 under a system prompt the client edits in the admin; the answer is stored as a post of its own and shown back to the visitor, who can then leave an email address to receive it. Written to the WordPress Coding Standards, with automated checks and tests on every change.
In both, the API key and the prompt live in the site’s settings, not in the code.
Why it matters
The model does one bounded task, its output lands in ordinary WordPress data, and editors can tune its behavior without a developer.