What is a Signal?
A Signal is a structured, source-based record of a documented change in real work: a task, workflow, role, workforce response, or capability requirement. The Tracker is a curated evidence explorer, not an AI news feed, adoption ranking, forecast, or representative survey.
Inclusion rule: WHO / WHAT / HOW
A candidate enters the Tracker when the evidence identifies WHO is changing, WHAT work is changing, and HOW the change is being carried out. Claims that describe only an AI product, intention, or broad investment do not by themselves establish a workforce Signal.
Core Taxonomy v1.0
The primary Work Transition describes the main observed change. Secondary Work Transitions capture additional documented changes without replacing the primary classification.
Evidence ladder
E1 Reported is a reported or prospective claim and is not presented as a verified public Tracker Signal. E2 Operational documents a system in use. E3 Behavioural documents changed behaviour or practice. E4 Outcome documents an explicit outcome claim. The public threshold is normally E2 or higher.
Not Yet Evidenced
When a Signal does not contain an evidenced workforce response, the Explorer labels it “Not yet evidenced.” This is an explicit evidence state, not a conclusion that no response exists.
Sources, facts, and interpretation
Source type describes where evidence comes from; it is not the same as evidence strength. “What changed” and observed capability shifts are factual fields. Workforce implication, decision relevance, and capability implications are clearly marked Observatory interpretation.
Organization identity and independence
Company names and trademarks are used solely to identify organizations referenced in independent evidence records. Inclusion does not imply endorsement, sponsorship or partnership with AIGW Policy Observatory. All trademarks remain the property of their respective owners.
Evidence boundaries
The Tracker does not infer layoffs from automation. Hire and Reduce require direct workforce-response evidence. Duplicate or overlapping transition episodes are handled as research-review decisions rather than inferred from similar headlines.
Update cadence and limitations
The Tracker is updated monthly. The current evidence collection is curated and should not be interpreted as a statistically representative global sample. Absence of a Signal is not evidence that a transition is absent.