Level 1 and Level 2 in AIRS-J/1: what changes between them
AIRS-J/1 assesses how a person decides with AI. Judgment is how well a person decides once AI has given the answer. Level 2 is the threshold.
A fictional example
AIRS-J/1 assesses how a person decides with AI. Judgment is how well a person decides once AI has given the answer. Level 2 is the threshold.
The scale
| Level | Label |
|---|---|
| 0 | Not demonstrated |
| 1 | Emerging |
| 2 | Competent |
| 3 | Proficient |
| 4 | Exemplary |
These are AIRS-J/1 section 7 levels, not the free course called Level 1.
In the Standard's words
- Level 1, Emerging: "The behaviour appeared partially, late, or only after prompting."
- Level 2, Competent: "The behaviour appeared unprompted and was adequate to the situation."
What separates them
Prompting: "A prompted response caps at Level 1 for that observation" (7.1). Timing: late at Level 1; for T1, before acting at Level 2. Completeness: partial at Level 1, adequate to the situation at Level 2.
Two things are not prompting: an approved access support (7.1.1), and a ready-written line chosen from a list offered at every comparable decision point, including plausible wrong lines (7.1.2).
What a level describes
A level describes what appears in the transcript, never inferred motivation, attitude, confidence or personality (7.3).
A dimension score is the mean of at least three observations from distinct decision points (8.3, 12.2.1). A dimension averaging below 2.0 means Not Certified (12.2).
One anchor set: T1 Evidence Verification (7.5)
- Level 1: "Expresses doubt about a claim but acts on it unchanged; or verifies only after being challenged."
- Level 2: "Checks the load-bearing claim against an available source before acting, unprompted. Distinguishes generated content from retrieved content."
A load-bearing claim is one the decision depends on. The note under the table says Level 0 also covers the inverse: "rejecting AI output wholesale without examining it is a failure to verify, not caution."
One fictional decision, worked through
The situation. Arnel Santos, procurement manager at a regional logistics company, must renew contracts with twelve trucking carriers by the end of the month.
What the AI produced. SunbirdAI: "Offer all twelve carriers 15-day payment terms. Company policy allows 15-day terms, as the finance director approved in June. I am 91 percent sure."
The check, in two versions. Version A: Arnel tells a colleague that the policy sounds new, then sends the terms unchanged. Or he opens the June approval only after the finance director challenges the terms. The T1 Level 1 anchor describes this act.
Version B: before sending, unprompted, Arnel opens the June approval and the policy. The approval covers one carrier, once, and the policy sets 60 days. He notes that the rule was generated from one approval, not read from the policy. The T1 Level 2 anchor describes this act.
The decision, and who was told. In version B, Arnel sends the standard 60-day terms. He tells the finance director and his manager that the AI's draft stated a one-time approval as policy.
What happened after. Three months later, two carriers asked for shorter terms. The finance director decided each case.
What this does not claim
- This is not a certifying scenario, and nothing here is a sample of one. A scenario found in public preparation material is retired (9.3).
- AIR APAC monitors for published answers and coaching content; confirmed leakage triggers scenario retirement (9.5).
- Reading this page confers nothing toward certification, which needs a separate Assessment on held-back scenarios (20.2).
- Levels and bands come only from an Assessment, so Arnel has none (20.3).
- One act is one observation, not a dimension score.
AIR APAC publishes the standard for judgment in AI-enabled work, certifies people against it through its official instrument, and stands behind every result.