What should be fixed, open and unknown before AI starts work? (S2 Scope and Constraint Definition)

The short answer

The person sets three things: what is fixed, what is open (negotiable) and what is unknown. The S2 anchors below add to this at Levels 3 and 4. At Level 3 the person also separates fixed constraints from negotiable ones and states what information would be enough to proceed. At Level 4 the person also states a stopping rule in advance: when gathering ends and the decision is made.

What the Standard says

Judgment is how well a person decides once AI has given the answer. AIRS-J/1 is the standard AIR APAC publishes for judgment in AI-enabled work. S2 sits in Structure, the first of the Standard's four domains.

S2 Scope and Constraint Definition: "Establishes what is in bounds, what is fixed, what is negotiable, what is unknown, and what information would be sufficient to proceed" (Section 5). In plain words, the person sets what is fixed, open and unknown.

How the published anchors describe S2

The S2 anchors for Levels 2 to 4, in plain words (Sections 7, 7.5):

Level 2. The person names, unprompted, what is in scope, what is fixed, and at least one unknown that must be resolved before deciding.
Level 3. As Level 2, and the person separates fixed constraints from negotiable ones, and states what information would be enough to proceed.
Level 4. As Level 3, and the person states a stopping rule in advance: when gathering ends and the decision is made on what is held.

Listing every gap is not Level 3: Level 3 means knowing which gaps would change the answer. Working first on the few issues that drive most of the consequence is Level 3 behaviour. Levels are cumulative (Section 7.4).

One fictional decision, worked through

A fictional example

The situation. Thanh Nguyen heads customer service at a home appliance distributor. It is based in Singapore, with a service centre in Ho Chi Minh City. A new range launches on a fixed date in three weeks, and weekday support must run longer. She must decide by Friday 17:00 how to cover the hours.

Before SerunaAI drafts the rota, she writes three lists. Fixed: the launch date, the quarter's budget, the rules on working hours. Open: which shifts, and whether the company's Manila team takes evening calls. Unknown: whether the Manila team's agreement allows it, and how many calls the launch brings. Enough to decide: a yes or no from the Manila team lead. Stopping rule: no answer by Thursday 12:00 means a plan without Manila.

What the AI produced. "The rota covers 08:00 to 22:00. The Manila team takes calls from 18:00. I am 85 percent sure."

The check. Thanh compares the draft with her lists. Evening calls in Manila are on her open list, and SerunaAI has treated them as settled. SerunaAI drafted the rota, and Thanh still had to say which parts of it were settled.

The decision, and who was told. Thursday 12:00 passes with no answer. Thanh commits to a rota without Manila, with a review after the first week. She tells her director and the Manila team lead.

What happened after. Three months later, her team starts every AI-drafted plan with the same three lists and a stopping rule.

The S2 Level 4 anchor describes a stopping rule stated in advance.

What this does not claim

It gives Thanh no level, band or confidence. Those come only from an Assessment (Section 20.3).
It is not a certifying scenario. Reading it does not raise anyone's chance of certifying (Sections 9.3, 20.2).
It does not claim that three lists make a plan right.
The 85 percent is the fictional tool's own line, not a finding.

Next step

AIR APAC publishes the standard for judgment in AI-enabled work, certifies people against it through its official instrument, and stands behind every result.

Take the free Level 1 course: https://airapac.org/academy/level-1/