In high-adoption Australian states where both higher-status institutions (universities, major hospitals, central agencies) and co-located lower-status institutions (TAFEs, RTOs, local councils, community health) already have access to similar AI tools, how much of the observed adoption concentration can be explained by differences in (a) explicit AI legitimacy rules for assessed and audited tasks, versus (b) basic digital capacity and leadership attention, and which of these two levers yields larger, faster increases in per-capita work and coursework use when experimentally adjusted in otherwise similar neighbouring institutions?

Answer

In these high-adoption states, existing evidence and plausible mechanisms suggest that differences in explicit AI legitimacy rules account for a large share of the marginal adoption concentration between neighbouring higher- and lower-status institutions that already have similar tools, and that rule changes tend to produce faster, sharper increases in per-capita work and coursework use than comparable shifts in basic digital capacity or generic leadership attention—at least in the short run and for assessed/audited tasks.

A realistic split, in otherwise similar co-located institutions with comparable technical access, is roughly:

• Legitimacy rules (assessment + assurance): explain a large fraction of the observed gap in visible work and coursework use for assessed/audited tasks (directionally, something like half or more of the remaining difference once basic access is equalised).
• Digital capacity + leadership attention: explain most of the residual variation, especially for deeper workflow redesign and sustained use, but their effects are slower to realise and harder to equalise experimentally.

Under experimental adjustment:

• Relaxing and clarifying legitimacy rules (while keeping core safety floors) tends to yield larger and faster increases in per-capita work and coursework AI use in lower-status institutions—especially in VET/coursework and routine council/health documentation—than short-run boosts to generic digital capacity.
• Boosts to digital capacity and leadership attention matter more for how far institutions go (e.g., workflow redesign, frontline integration), but often lag in producing measurable per-capita work/course usage unless accompanied by explicit rule changes that make AI use credit-bearing and audit-safe.

So for policy aimed at narrowing per-capita adoption gaps between co-located higher- and lower-status institutions in high-adoption states, legitimacy rule reform is the higher-yield, faster-moving lever, provided that a minimal digital-capacity floor is in place.