Why does a standard for judgment in AI-enabled work exist?
AI systems now produce the answer. A person still has to decide whether to act on it. AIRS-J/1 was written so that the quality of that decision can be read and compared, by published rules.
Judgment is how well a person decides once AI has given the answer. AIR APAC publishes AIRS-J/1, the Standard for Judgment in AI-Enabled Work, to describe that skill and to certify it. This article sets out the problem the Standard was written to solve, and what it gives that a course record or a confidence rating cannot.
The problem
The Standard states the problem in two sentences: "AI systems now produce the answer. They do not carry the consequence of acting on it" (Section 1.1).
The work of a person has moved. The human contribution now lies in deciding "whether work is fit to act on" and "who is entitled to be told". It also lies in deciding "what may not be delegated, and what happens when it turns out to be wrong" (Section 1.1).
The Standard then names two gaps. Organisations have no common language for that contribution. They also have no way to compare it "across two people, two teams or two markets". It then says: "Training attendance, tool proficiency certificates and self-reported confidence do not measure it" (Section 1.1). The AIR APAC article Your AI training records prove nothing about judgment sets out that case.
What a standard adds
Section 1 of AIRS-J/1 gives four reasons for the Standard. In plain words:
- A credential that travels. A certified person holds "a portable credential whose meaning does not depend on their employer". The credential belongs to the person and stays with them after a change of employer (Section 14).
- Results that compare. Two people assessed in different organisations, countries or delivery settings receive comparable results.
- Criteria anyone can inspect. A candidate, an employer, a regulator or a competitor can read the criteria against which a decision was made. The rubric is published in full.
- A reading an employer can rely on. An employer can rely on "an expert, independent reading of a person's judgment" from an employment-grade result (Section 12.8). The employer uses it as one input (Section 12.6).
A course record shows that a person completed a course. A confidence rating shows how sure the person feels. Neither is a reading of how the person decided when an AI answer was in front of them.
The Standard also keeps training and certification apart. A candidate may present for assessment without prior training. Certification outcomes "are decided by the rules in Section 12 alone, never by what a candidate or employer bought" (Section 4.3).
The rule that governs the design
AIRS-J/1 names one constraint above the others: "Comparability is the governing design constraint" (Section 1). Where customisation and comparability conflict, the Standard keeps the result comparable. Where comparability and evidential honesty conflict, it chooses honesty: "a result that cannot be defended is worse than no result".
A result is built to be comparable across the organisations that receive one. Where the evidence is too thin, no certification is issued. Certification needs Moderate or High confidence. Confidence describes how much evidence there is, not how able the person is.
Why managers need it
The last paragraph of Section 1.1 is about managers. It says: "Managers are often reluctant to judge their people's judgment, because the consequences of getting it wrong fall on them". The Standard gives them "a published, evidence-based reading to rely on".
That reading is one input. Under Section 12.6, an employment-grade result (a result at Moderate or High confidence) is AIR APAC's expert reading, "which the employer may rely on as one input". The Standard is plain about who decides: "The employment decision is the employer's" (Section 2.1).
What it covers, and what it does not
Section 2 of the Standard gives the scope in one sentence: "This Standard assesses how a person decides, not what they know". It does not assess skill with any AI tool, or knowledge of a professional field.
One fictional decision, worked through
A fictional example. The company, the people, the AI tool and every number are invented.
The situation. Rohit Malhotra is head of operations at a regional freight forwarder. Its head office is in Singapore, and it has offices in Kuala Lumpur and Jakarta. The firm uses CanasteroAI to draft quotes for repeat customers. Rohit has two weeks to decide which of his two team leads may approve AI-drafted quotes up to a higher limit without a second reviewer. The year-end volume rises after that. Both team leads completed the same AI course. Both rate themselves "very confident".
What the AI produced. That week, CanasteroAI drafted the same kind of quote in both offices: "Accept the quote at SGD 18,400. It matches this customer's last three orders on the same lane. I am 90 percent sure."
The check. Rohit sets the course certificates and the confidence ratings aside. He opens the approval notes for the last twenty quotes from each office. Qistina Azman, the Kuala Lumpur team lead, records a basis for each one. She checked the lane rate against the carrier's current tariff. On two quotes she noted a changed fuel surcharge and adjusted the price. The Jakarta notes say "agreed with AI" each time. The customer accepted every quote in both offices, so the outcomes do not tell the two team leads apart.
The decision, and who was told. Rohit raises the approval limit for the Kuala Lumpur team. He keeps the second reviewer in Jakarta. He asks that every approval note state what was checked. He writes to both team leads and to the commercial director with his reasons.
What happened after. Three months later, the Jakarta notes state a basis. Rohit raises the Jakarta limit too. The approval form keeps one line: "What did you check, and against what?"
What the example shows, and where it stops
Rohit could read the notes of two offices in one company. A standard does the same work with published criteria, so that the reading is comparable across organisations and stays with the person. The parallel lies in the method, not in the evidence. AIRS-J/1 does not assess approval notes or any other workplace record. It reads what a person does in an Assessment, against published rules.
What this does not claim
- It does not claim that certification would have solved Rohit's problem, or that the company in the example used the Standard.
- It does not claim that a workplace record, an approval log or a quote is assessed under AIRS-J/1.
- It does not claim that a result settles any decision about a job. The employment decision stays with the employer (Section 12.6).
- It does not claim that training has no value, only that attendance does not measure this skill (Section 1.1).
- It does not claim that any law or regulator requires the Standard.
AIR APAC publishes the standard for judgment in AI-enabled work, certifies people against it through its official instrument, and stands behind every result.