AI Fluent HR

Methodology

Watch-and-quiz doesn't prove capability. We grade what you actually produce.

Most "AI for HR" courses sell hours of video and a certificate of completion. We sell evidence: rubric-scored artefacts you wrote, defended, and shipped. Here's how the engine works.

The 4D Framework

Discover, Develop, Deliver, Defend. Every lesson, game, and Practice Bay scenario maps to one of these four durable competencies — not to a tool name that will be obsolete in 18 months.

Rubric-first design

Every scenario starts with the rubric, not the prompt. HR experts write the criteria (3–6 dimensions, 5-point scale) before we ever ship the scenario. You see the rubric before you submit.

AI grading, human calibration

Practice Bay uses an LLM to grade against the rubric. We sample ~10% of attempts and re-grade with an HR reviewer; we publish the inter-rater agreement quarterly and recalibrate prompts when it drops.

Skill Passport

Every passing artefact — quiz, exercise, Practice Bay — is logged to your passport with the score, the rubric, and a permalink. Your certificate lists artefacts, not hours.

POPIA-grade by default

No learner PII is sent to the LLM. Every call is logged. Tenants are strictly isolated. Org admins see what their team produced; we don't see their content.

Tracks, not catalogs

A track is a curated sequence — 6–12 lessons + Practice Bay milestones + a capstone — leading to a verifiable certificate. No filler. No 40-hour modules padded with introductory video.

How a scenario gets onto the platform

  1. An HR practitioner proposes a real-world artefact (a JD, a feedback line, a comp memo).
  2. A rubric is drafted with 3–6 dimensions and example anchors at 1, 3, and 5.
  3. Three expert reviewers grade five sample submissions blind; we only ship if inter-rater agreement is ≥ 0.7 (Krippendorff's α).
  4. The LLM grading prompt is tuned against the same five samples until model-vs-human agreement crosses the same threshold.
  5. We publish, monitor weekly, and recalibrate if scores drift.

What we won't do

  • We won't sell a "ChatGPT for HR" course that's just screenshots from one vendor's UI.
  • We won't issue a certificate for watching videos.
  • We won't put learner PII into a third-party model.
  • We won't claim our AI grader is infallible. It's calibrated, sampled, and auditable.