A public standard for human judgment in work done with AI

AIRS-J/1: the Standard for Judgment in AI-Enabled Work

AIR APAC defines what good judgment in AI-enabled work looks like, assesses people against it, and stands behind every result. We publish the rules, give the reasons for every level, and correct a result when an appeal succeeds. Each person is assessed on what they actually do in a realistic work scenario, not on what they say about themselves.

Assess your team against the Standard Read the Standard (v2.3)

Download the Standard as a PDF

Status

In force. Certification is open.

AIRS-J/1 has three uses: developing people in workplace training, checking the readiness of final-year students and recent graduates, and informing employment decisions. An employer can pay for a person's assessment, and the person decides who sees the result. AIR APAC also certifies the trainers who run sessions on StepSim, the assessment simulator built by Exeter Labs.

Current version

AIRS-J/1 v2.3, in force, dated 2026-10-19

Certification

Open. Results come from the published rules only.

Official instrument

StepSim, built by Exeter Labs: a text-based work simulation that keeps a full record of every decision. Written rules read what the person did. No AI grades anyone.

Structure

Nine dimensions under four domains, scored 0 to 4 against published behavioural anchors

Credential

Valid two years, belongs to the person who earns it, listed in a free public register

Take an assessment

Assessments are arranged through your trainer or your organisation's learning and development team, who confirm your date.

The competency model

Nine dimensions, four domains.

The nine dimensions describe what a person does when they decide with AI. Each is scored from 0 to 4. Level 2 is the threshold for competent performance.

Structure

Framing the problem before reaching for a tool.

S1Problem Framing
S2Scope and Constraint Definition

Test

Checking the evidence, including evidence produced by AI.

T1Evidence Verification
T2Assumption and Alternative Testing
T3Risk and Failure Anticipation

Engage

Acting within your authority, and involving the right people.

E1Decision Rights and Escalation
E2Stakeholder Alignment and Disclosure

Perform

Deciding proportionately, and owning the decision.

P1Proportionate Commitment Under Uncertainty
P2Consequence Ownership and Correction

See all nine dimensions, with a worked example of each The nine dimensions in detail

AIR Academy programmes, informed by the Standard, help teams practise these nine dimensions. An optional assessment leads to certification. See all fifteen programmes.

Governance

Who governs the Standard

Fourteen named people sit on the Review Board, unpaid, and review every major version before release. The Standard also names an External Academic Reviewer. AIR APAC, as the owner of the Standard, holds a veto over the board, and every dissent is recorded in the version history.

Review Board

Fourteen named members, unpaid, listed with their roles on the People page

External Academic Reviewer

Prof. Mudiarasan Kuppusamy, UNITAR International University

Owner's veto

AIR APAC decides and holds the veto. Every dissent from a released version is recorded.

Dissent recorded

None so far

Board membership is not an endorsement of the Standard or of any instrument.

Meet the Review Board How the Standard is governed

Participate

How to take part

Employers

Assess your team

Your organisation receives each result by name and a team view. Candidates are told before they start. The employment decision stays yours.

Talk to us about assessing your team

Trainers and partners

Deliver training

Deliver AIR Academy programmes to your clients, and become a certified StepSim facilitator. You keep the client relationship. No partner is ever paid according to candidate results.

Become a StepSim facilitator

Individuals

Get certified

Any individual may present, with or without prior training. The result is released only to the parties you name, and the credential is yours.

Students, graduates and institutions

Campus

For final-year students and recent graduates, on workplace scenarios at entry level, with rules for institutions (section 13). The holder is judgment-ready for AI-enabled work.

The Campus route

Anyone

Submit a correction

Found an error, or evidence against a claim? Tell us. Every submission is logged, and errors in a rule go to the Review Board.

Send a correction

What a result and a credential claim

Every claim names its basis.

Every claim about a credential must name its basis in the Standard. A credential says the holder met published rules for judgment in AI-enabled work. It does not measure intelligence, personality, AI technical skill or domain expertise. See what each claim needs and where each check stands.

Every assessment adds to the evidence. A public validation programme checks that the levels are stable, fair and meaningful, and publishes the results whether they are favourable or not.

Validation status, check by check

Reference

Read any version of the Standard

Version files are additive. v1.0 is never overwritten, and the changelog quotes the history row for each version.

In force

v2.3

Dated 2026-10-19. Nine dimensions, four domains, the anchors, thresholds and decision rules open to read, with the rules for training, Campus and employment use, the sponsored route, facilitator certification and guidance on deciding under uncertainty.

airapac.org/standard/v2-3

v2.2, v2.1, v2.0 and v1.1

Each kept at its own permanent URL and still readable.

airapac.org/standard/v2-2

airapac.org/standard/v2-1

airapac.org/standard/v2-0

airapac.org/standard/v1-1

First published

v1.0

The first published version, kept at its own permanent URL and still readable.

airapac.org/standard/v1-0

Everything about the Standard, in one place: how it is governed, how certification works, how to appeal, and how to check a credential.

Each page of the Standard, and what it carries.
Destination	What it carries	URL
Read the Standard	The full text of AIRS-J/1 v2.3, in force from 19 October 2026, with every level description, threshold and decision rule. A designed PDF is offered for download.	/standard/v2-3
The nine dimensions	The nine observable dimensions of judgment under four domains, with what each one reads and how it is scored.	/standard/dimensions
Governance	The named Review Board, the External Academic Reviewer, and the rules that bind AIR APAC itself.	/standard/governance
Instruments	The instruments that may produce a certifying result, and the declared interests behind them.	/standard/instruments
StepSim	The official instrument: a scenario simulation with a retained decision trace, specified in Annex F of the Standard.	/standard/instruments/stepsim
Certification	How certification works, what a candidate is assessed on, and the decision rules in section 12.	/standard/certification
Apply for an assessment	The application route for an assessment on StepSim.	/standard/certification/apply
Campus	The route for final-year students and recent graduates, on workplace scenarios at entry level, with the rules for institutions.	/standard/campus
Validation status	The eight validation checks in section 11.7, what each one checks, and where each stands today.	/standard/validation
The public register	Every certifying result is listed with the date it was awarded and the assessment it came from. Credentials are valid for two years. Anyone can check one, free.	/standard/register
Verify a credential	Verification by code, free and public. Registration details are checked against the register entry.	/standard/register/verify
Appeals	How to appeal a result or a decision, and the time limits that apply.	/standard/appeals
Complaints	How to complain about the conduct of an assessment, an assessor or a partner.	/standard/complaints
Access and support	Adjustments and support for a candidate who needs them, agreed before a sitting.	/standard/access-support
Version history	What changed between versions, with the history row quoted for each version.	/standard/changelog

Citation: AIR APAC. AIRS-J/1: Standard for Judgment in AI-Enabled Work, v2.3. https://airapac.org/standard/v2-3.

Each published version carries a SHA-256 digest, so you can check that the text has not changed.

Questions and worked examples.

Short pages that explain this part of the Standard in plain words. The Standard's own text, above and in AIRS-J/1 v2.3, governs.

Why does a standard for judgment in AI-enabled work exist? (Explainer)
Words used in the Standard, in plain English (Glossary)
What changed in version 2.1 of the Standard? (Guide)
What does the Standard measure, and what does it not measure? (Guide)
What is AIRS-J/1, in plain words? (Guide)
Which four domains does the Standard use? (Guide)
Who publishes the Standard, and who may use it? (Guide)