Read one summary instead of a hundred CVs
WP Career Board Pro scores every application against the job it was sent to, writes a one-line reason and a neutral summary of the candidate, and ranks them for you. It runs on Claude, OpenAI or a self-hosted Ollama model, so candidate data can stay on infrastructure you control.
Applications, ranked by fit
The problem
Posting the job was never the hard part. The hard part is the ninety applications that arrive in a week, most of them wrong for the role, all of them needing to be read by somebody before anyone gets a reply.
- Each application gets a fit score from 0 to 100, measured against that specific job
- A one-sentence reason explains the score, so you can disagree with it rather than trust it blindly
- A neutral one or two sentence summary of the candidate's background, for the ones you still want to read
- Applications rank by score, so the shortlist is the top of the list rather than an afternoon's work
- Scores are cached against the application, so reopening a job costs nothing and bills nothing
- Runs in batches, with the batch size adjustable for your host's limits
- Choose Claude, OpenAI or Ollama, and set the completion and embedding providers separately
- With Ollama it is self-hosted, so CVs are never sent to a third party at all
- Job and candidate embeddings power matching, not just scoring
Who it's for
Small teams hiring without a recruiter
Get to a shortlist in the time you have, which is usually an evening, not a working day.
Anyone holding CVs under GDPR
Point it at a self-hosted Ollama model and the candidate data never leaves your server.
Board operators with high volume
Rank hundreds of applications per role in batches instead of reading them in order of arrival.
Developers
Register your own provider through a filter rather than being locked to the three that ship.
Common questions
What exactly does the AI produce for each application?
Three things, from a single model call: a fit score as an integer from 0 to 100, a one-sentence reason for that score, and a neutral one or two sentence summary of the candidate's background. All three are stored against the application, so they are there next time you open it and you are not paying to generate them twice.
Can I keep candidate CVs off third-party servers?
Yes. Ollama is one of the three supported providers and it runs on your own machine, defaulting to http://localhost:11434 with llama3 for scoring and nomic-embed-text for embeddings. Nothing is sent anywhere. That is the configuration to choose if you are handling CVs under GDPR and would rather not add a processor.
Which AI providers are supported?
Anthropic's Claude, OpenAI, and Ollama for self-hosting. The completion provider and the embedding provider are set separately, so you can score with one and embed with another. Developers can register additional drivers through the wcbp_ai_provider_drivers filter, so the three that ship are a starting point rather than a ceiling.
Does it decide who to hire?
No, and it should not. It ranks and summarises so a person reads the right applications first. The one-line reason exists precisely so you can see why something scored well and disagree with it. Treat the score as a reading order, not a decision, and keep a human on every rejection.
What does it cost to run?
On Claude or OpenAI you pay that provider for the calls, and results are cached per application so the same CV is not scored twice. On Ollama it costs you nothing beyond the machine it runs on. Batch size is adjustable through a filter, which matters on shared hosting where a long run would otherwise time out.
Is this in the free version?
No, AI screening is part of Pro. The free version handles the whole board: publishing jobs, taking applications, employer and candidate dashboards, search, moderation, GDPR export and erase, and the JobPosting schema that makes roles eligible for Google's job results. Pro adds the screening, the pipeline, resume profiles and monetisation.
How does matching differ from scoring?
Scoring compares one application to the job it was sent to. Matching uses embeddings generated for jobs and candidates to suggest people who never applied to that particular role. Scoring sorts your inbox; matching finds someone already in your database who fits.
Run it on your own WordPress.
Download the free plugin, run the 5-minute wizard, and turn this on. Your board, your data, no SaaS rent.