The words in AIRS-J/1, defined plainly

AIRS-J/1 assesses how a person decides when working with, alongside or in supervision of AI systems. It does not assess what a person knows: not skill with an AI tool, not domain knowledge, not intelligence and not personality (Sections 2, 2.1). Judgment is how well a person decides once AI has given the answer.

This glossary follows AIRS-J/1 v2.3, in force 19 October 2026. It defines 34 words. Each entry gives the sections of the Standard that the word comes from, then a plain reading. Where an entry and the Standard differ, the Standard governs.

Access support

Sections of the Standard. 3, 7.1.1, 12.6.

Plain reading. An approved change to presentation, timing or the way a candidate answers. It removes a barrier that is not part of what is assessed. It never reveals a decision path or an expected response, so it carries no level cap. The employer is never told whether an access support was used.

Anchor

Sections of the Standard. 3, 7.3, 7.5.

Plain reading. The behavioural description that defines one level, from 0 to 4, on one dimension. Each of the nine dimensions has its own anchor set. Anchors describe what appears in the transcript. An assessor may not score on inferred motivation, attitude, confidence or personality.

Appeal

Sections of the Standard. 18.5, 18.6.

Plain reading. A request by a candidate for review of a certification-status decision, an employment-grade result released to an employer, an access-support decision, or a suspension. It must be made within 30 days. The first appeal is free. Someone who was not involved in the original decision or delivery reviews it. A development reading decides nothing and cannot be appealed.

Assessment and session

Sections of the Standard. 8.4, 12.6, 20.3.

Plain reading. This is a plain reading, as Section 3 does not define the capital A. An Assessment is a sitting under the certification rules, on held-back scenarios. Levels per dimension and bands come only from an Assessment. A session is practice or training and gives a development reading. Before a session, people are told a hiring or certification decision uses a separate Assessment.

Band

Sections of the Standard. 12.4.

Plain reading. The way a certified result is reported, in place of a raw number. Certified: composite 2.3 to 2.9. Certified with Merit: 3.0 to 3.4, with no dimension below 2.5. Certified with Distinction: 3.5 and above, with no dimension below 3.0. Bands prevent false precision and ranked lists of people.

Campus track

Sections of the Standard. 13, 13.1.

Plain reading. The track for final-year students of institutes of higher learning, and for graduates in their first two years of work. Scenarios are entry-level workplace decisions, with a named supervisor to escalate to. A Campus certification is not equivalent to a Professional certification and cannot be converted into one.

Certified function

Sections of the Standard. 2.2, 12.5.

Plain reading. What a credential certifies: the exercise of judgment in AI-enabled work, using domain content supplied within an approved assessment. It does not certify occupational competence, domain expertise, comprehensive AI literacy, management of people, or performance in an unbounded workplace. Every report and credential carries this scope in plain language.

Composite

Sections of the Standard. 12.2.1, 12.3, 12.4.

Plain reading. The arithmetic mean of all nine dimension scores. To certify, the composite must be at or above 2.3. A profile at exactly 2.0 on every dimension gives a composite of 2.0, so it does not certify. Results are reported in bands, not as raw composites.

Confidence

Sections of the Standard. 10.

Plain reading. A statement of how much evidence stands behind a result. It describes the evidence, not the ability of the candidate. There are four grades: High, Moderate, Provisional and Insufficient. Certification is issued at Moderate or High only. Confidence is printed on the credential and shown in the register.

Consequence family

Sections of the Standard. 6.1, 6.2.

Plain reading. A kind of harm a decision can cause. There are six families, C1 to C6:

people safety and welfare
financial and fiduciary loss
operational and service continuity
information and privacy harm
trust, reputation and legitimacy
environmental and community-system harm

Families are recorded so that decisions with similar consequences are compared with each other. They are not scored.

Credential code

Sections of the Standard. 13.4, 14.1, 14.2.

Plain reading. The code that identifies one credential in the register. Verification by credential code is the default: any party that holds the code can verify the entry. A Campus holder, for example, releases a credential to an employer by giving the code or by naming the party.

Decision point

Sections of the Standard. 3, 8.3.

Plain reading. A moment in an assessment at which the candidate must commit to an action, a position or a piece of work that carries a cost. Each observation comes from one decision point. Two observations from the same decision point count as one observation.

Development reading

Sections of the Standard. 3, 10.1, 12.8, 20.3.

Plain reading. A reading from practice, a training session, or a result below Moderate confidence. It shows the decisions of a person and how the published rules read them. It carries no band, and a practice or training reading carries no level. It decides nothing about employment or certification, and it is never released to an employer in a hiring context.

Dimension

Sections of the Standard. 3, 5, 12.2, 12.5.

Plain reading. One of the nine judgment competencies the Standard assesses, for example T1 Evidence Verification or E1 Decision Rights and Escalation. Every candidate receives a score for each dimension. To certify, no dimension may be below 2.0.

Domain

Sections of the Standard. 3, 5, 12.2, 12.2.1.

Plain reading. One of four groups of dimensions: Structure (S1, S2), Test (T1, T2, T3), Engage (E1, E2) and Perform (P1, P2). A domain score is the mean of its dimension scores. Each domain must be at or above 2.0 to certify.

Employment-grade result

Sections of the Standard. 3, 12.6, 12.8.

Plain reading. An Assessment result from an approved instrument at Moderate or High confidence, or a valid credential shown by its holder. Only this kind of result may be used as the reading of AIR APAC in a hiring, promotion or appraisal decision. The result is one input. The employment decision is the employer's.

Held-back scenario

Sections of the Standard. 8.4, 9.3.

Plain reading. A certifying scenario kept out of every demonstration and every practice while it is in use for certification. The official instrument uses at least three held-back Assessment scenarios. A scenario is retired from certification when it is exposed too often or appears in public preparation material. A retired scenario may then be released for practice.

Level

Sections of the Standard. 7, 7.2, 7.4.

Plain reading. The score of one observation, from 0 to 4: 0 Not demonstrated; 1 Emerging; 2 Competent, which is the threshold; 3 Proficient; 4 Exemplary. Levels are cumulative, so each level needs everything in the level below. Level 4 is not seniority, and it is open on every track.

Model-assisted proportion

Sections of the Standard. 11.4, 12.5.

Plain reading. The share of observations in which a model proposed a level and a qualified human assessor confirmed or overrode it, with reasons recorded. A score from a model never stands alone as an observation. The proportion is recorded on the result, reported to the candidate and disclosed in the register entry.

Near-cut rule

Sections of the Standard. 12.8.1.

Plain reading. Before an employment-grade result is released for hiring or promotion, AIR APAC checks one thing. Would a change of one level in any single observation move the outcome between Certified and Not Certified? If so, the candidate is offered one further held-back mission on an unseen form, at no charge. The released result states whether the rule was applied.

Non-compensatory

Sections of the Standard. 12.1, 12.2, 12.3.

Plain reading. Strength in one domain does not offset weakness in another. The composite is one of six rules, and the dimension and domain floors still apply. A candidate with three high dimensions cannot certify while at 1.0 on another dimension.

Observation

Sections of the Standard. 3, 8.3, 8.8.

Plain reading. A single scored instance of one dimension, from one decision point, supported by identified evidence in the transcript. Each dimension needs at least three independent observations. An observation whose evidence cannot be pointed at does not exist.

Official instrument

Built by Exeter Labs.

Sections of the Standard. 4.9, 8.1, 16.3.

Plain reading. StepSim is the official instrument of AIRS-J/1: a scenario simulation that keeps a record of every decision. The instrument applies the rubric; it does not define it. Exeter Labs builds and operates StepSim, AIR APAC's official assessment instrument. The owner of AIR APAC also owns Exeter Labs.

Practitioner track

Sections of the Standard. 13.

Plain reading. The track for those who lead AI adoption or assess others. Its scenarios add conflict between several parties, a contested mandate, and doubt about who owns the decision. The dimensions, anchors, thresholds and decision rules are the same as on every other track.

Professional track

Sections of the Standard. 13, 13.2.

Plain reading. The track for practising managers and individual contributors. Its scenarios carry the full range of stakeholders, consequences for the organisation, and live limits of authority. An employer that commissions an assessment of its own employees uses the Professional track or a role edition of it.

Prompting

Sections of the Standard. 7.1, 7.1.1, 7.1.2.

Plain reading. Help from the instrument that is directed toward the behaviour being assessed. The level of help is recorded in the transcript. A prompted response caps at Level 1 for that observation. An access support is not prompting. Choosing a line from a response bank that meets Section 7.1.2 is not prompting either.

Register

Sections of the Standard. 14.1.

Plain reading. The record AIR APAC keeps of credentials. It shows credential code, track, role edition where used, band, confidence, standard version, issue and expiry dates, status, and the scope in plain language. It does not show the employer, dimension scores, stakes profiles or any transcript content.

Role edition

Sections of the Standard. 13.3.

Plain reading. A version of a track whose scenarios are set in the work of one role family, for example finance operations or public service. It uses the same dimensions, anchors, thresholds and decision rules. A result from a role edition names the edition on the report, the credential and the register entry.

Stakes profile

Sections of the Standard. 3, 6.1, 6.4, 6.6.

Plain reading. The consequence families in play in one assessment instance: one primary family and at most one secondary. A certifying assessment must span at least three primary families, and at least one must involve irreversible or non-financial harm. The candidate report states the stakes profiles assessed.

STEP

Sections of the Standard. 5, 5.1.

Plain reading. The name of the competency model, taken from its four domains: Structure, Test, Engage and Perform. Nine dimensions is a reporting hypothesis, not an established structure. A dimension that does not separate from another in its domain is merged in the next major version, and the Standard says so.

Team view

Sections of the Standard. 12.7.

Plain reading. Counts of levels and bands across the people an organisation assessed, for staff who are not entitled to named results. A team view covers ten people or more; below ten, none is given. For development use, it counts decisions and how the rules read them, not levels or bands. AIR APAC publishes no ranked lists of people.

Transcript

Sections of the Standard. 3, 14.4, 14.5.

Plain reading. The complete retained evidence record of an assessment instance. Every observation must be derivable from the transcript alone. The transcript is the record AIR APAC keeps of the assessment. A candidate may obtain their own transcript and result, and may ask for factual errors to be corrected.

Validity

Sections of the Standard. 18.1, 18.2, 18.3.

Plain reading. A credential is valid for two years from issue. Recertification is a full reassessment against the version current at that time. A prior band gives no credit, no exemption and no reduced evidence requirement. A credential issued under an earlier version stays valid to its expiry and shows its version.

Verification by code

Sections of the Standard. 14.2, 14.3.

Plain reading. The default way to check a credential. Any party that holds the credential code can verify the register entry. Verification is free and needs no account, no registration and no commercial relationship with AIR APAC. A name appears in the searchable register only with the separate, revocable consent of the holder.

What this glossary does not claim

It adds no rule. Every entry is a plain reading of AIRS-J/1 v2.3, and the Standard governs.
Where the wording of the Standard is undefined or inconsistent, this glossary gives a plain reading and refers the point for formal correction of the Standard. A glossary cannot change a rule.
A result describes the judgment of a person in AI-enabled work at the level certified. It is not a finding about a whole job, and the employment decision stays with the employer (Sections 2.2, 12.6).
A credential authorises no one to act beyond their existing professional or legal authority (Section 2.1).

Next step

Read the certification rules