Skip to content
Docs

AI Features for Employers

Pro feature surface. Employers on Free still get the full job-posting form, the full applications screen, and standard candidate review. The AI-assisted shortcuts below appear when the site has Pro active and the relevant provider configured.

This page covers what employers can do with AI in 1.4.3: the Job Description Writer during posting, and AI applicant ranking with fit scores, reasons, and TL;DR summaries on the Employer Dashboard.

AI Job Description Writer

Section titled “AI Job Description Writer”

A button on the post-a-job form that turns a few form fields into a job description draft. Employers always review and edit the output before publishing - nothing is auto-posted.

Where it lives

Section titled “Where it lives”

On the Post a Job form (Employer Dashboard or /post-a-job/), above the description editor:

  • A “Generate with AI” button (icon + label). Present on both the full job form and the simple job form.
  • Visible only when:
    • Pro is active.
    • The wcb_ai_description_enabled filter resolves to true (Pro’s AiModule::is_enabled() returns true when an analysis or embedding provider is configured).
    • The employer’s role has wcb_post_jobs.

What gets sent to the AI

Section titled “What gets sent to the AI”

The writer endpoint (POST /wcb/v1/jobs/ai-description) accepts three fields:

Field Source on the form
title Job title input
company_type Company type / industry input
location Job location input

These are composed into a prompt asking the model to write a compelling job description (role overview, responsibilities, requirements) and to return clean semantic HTML (<h3>, <p>, <ul><li>), no markdown or code fences.

The flow

Section titled “The flow”
  1. Fill in the basics: job title, company, location, type.
  2. Click “Generate with AI.” Spinner appears for 5-15 seconds (depends on provider).
  3. A structured-HTML draft appears in the editor.
  4. Edit it. Add company-specific details, tone, anything the AI missed. The output is a starting point.
  5. Submit the job as normal.

Tips for employers

Section titled “Tips for employers”
  • A good title produces a better description. “Senior Backend Engineer (Rust + Tokio, billing service)” produces a more specific draft than “Software Engineer.”
  • The output is generic by default. AI doesn’t know your company. Plan to edit at least 30% of the draft for accuracy and tone.
  • Run it through your own eye before publishing. First-pass drafts often miss specific perks, your diversity statement, or HR legal language. Treat it like a junior intern’s first draft.
  • Each generation costs API credits (~$0.003 on OpenAI’s pricing per generation). Regenerating to get the tone right is still cheap.

Limits

Section titled “Limits”
  • Output is provider-determined - no per-request token cap in the plugin code.
  • Provider quality matters. Claude (Sonnet) writes more naturally; OpenAI is a bit more formal; Ollama / llama3 is the weakest writer.
  • Rate limit: 30 AI calls per user per hour, shared across all AI features.

AI applicant ranking

Section titled “AI applicant ranking”

Pro scores each application against its job (0-100 fit + a one-line reason + a neutral TL;DR summary) and surfaces it right on the Employer Dashboard - no custom code required.

Where it lives

Section titled “Where it lives”

On the Employer Dashboard applications view, when an analysis provider is configured (the wcb_ai_ranking_available filter is true):

  • A “Rank by AI fit” control sorts the loaded applications best-first.
  • Each applicant row shows an AI fit score (e.g. “87%”), the reason, and a TL;DR summary of the candidate’s background.
  • The detail view shows the same fit, reason, and summary.

Under the hood the dashboard calls GET /wcb/v1/ai/ranked-applications/{job_id}, which returns each application’s {application_id, score, reason, summary} sorted by score.

Caching - you are not re-billed

Section titled “Caching - you are not re-billed”

Fit score, reason, and summary are cached per application in post meta (_wcbp_ai_fit_score, _wcbp_ai_fit_reason, _wcbp_ai_summary, _wcbp_ai_scored_at). Re-opening the dashboard reuses cached values; ranking only computes applications that have never been scored. A force re-score is available via AiModule::score_application( $id, true ) from a custom integration.

Auto-score on submit (optional)

Section titled “Auto-score on submit (optional)”

Under Settings -> AI Settings, enable “Auto-score applicants on submit” (wcbp_ai_auto_rank). When on, Pro hooks wcb_application_submitted and schedules a background wcbp_ai_score_application cron event ~30 seconds after each new application, so the dashboard shows AI fit instantly without anyone clicking “Rank by AI fit.” It’s skipped when no analysis provider is configured.

What the score means

Section titled “What the score means”
  • 80-100 - strong match. Resume / profile genuinely aligns with the job. Worth interviewing.
  • 60-79 - partial match. Some required skills present, others missing. Read in full before deciding.
  • 40-59 - weak match. Limited overlap. Useful as triage signal, not as a filter.
  • Below 40 - minimal match. Most fields aren’t aligned.

The score is one signal, not the answer. Culture fit, location, salary expectations, and soft skills are not captured.

How it’s computed

Section titled “How it’s computed”

Pro sends the job title and the candidate’s resume text to the analysis provider with a JSON prompt asking for {score, reason, summary}. The candidate’s resume text comes from the Pro Resume module via the wcbp_candidate_resume_data filter (wired in Pro core - no add-on required). The model reply is parsed even when it’s wrapped in code fences, so scores are real rather than always zero.

Important caveats

Section titled “Important caveats”
  • Don’t auto-reject by score. AI bias is real; a low score on a great candidate (e.g. a career-changer) shouldn’t short-circuit human review. Pro never auto-rejects.
  • Sparse resumes score low. A candidate who hasn’t filled in the Resume Builder gives the AI little to work with.
  • Re-score is on demand. Cached scores stay until you force a recompute or auto-score handles a brand-new application.

Privacy considerations

Section titled “Privacy considerations”

Ranking is the most data-heavy AI feature - each applicant’s resume text is sent to the analysis provider on first scoring (then cached). If your hiring policy doesn’t allow that, set the analysis provider to Ollama to keep everything on your server.

What employers get without Pro

Section titled “What employers get without Pro”
Capability Free Pro 1.4.3
Post a job Yes Yes
Manual job description editor Yes Yes
AI Description Writer button No Yes
Review applicants Yes Yes
Sort by date, status Yes Yes
AI fit score + reason on the dashboard No Yes
AI TL;DR applicant summaries No Yes
One-click “Rank by AI fit” No Yes
Auto-score applicants on submit No Yes
Application pipeline (Kanban) No Yes
Bulk actions on applications Yes Yes

Where to go next

Section titled “Where to go next”