Among Australian public-sector organisations outside major capitals, how do different deployment models for shared AI capability—centralised state or Commonwealth platforms, regional consortia (e.g., multiple councils or health services sharing a service), or embedded AI coaches within existing agencies—compare in their effect on per-capita, work-related AI usage and on reducing adoption concentration between hubs and non-hub regions over a 3–5 year period?

Answer

Directional comparison over 3–5 years, assuming similar base connectivity and basic tools are available.

1. Centralised state/Commonwealth platforms

• Effect on per-capita work use (non-capital public sector):
  • Moderate. Low marginal cost and easy onboarding once set up, but weak local change management.
  • Typical pattern: a minority of regional agencies use the platform heavily; many register but use lightly.
• Effect on adoption concentration (hubs vs non-hubs):
  • Small reduction. Hubs still dominate usage; regions gain access but not necessarily depth.
• Best use: common, low-risk utilities (summarisation, drafting, search) integrated into existing state systems.

2. Regional consortia (shared services across councils/health/TAFEs)

• Effect on per-capita work use:
  • Higher than centralised alone when there is a funded lead and shared staff (e.g., a small AI/service team for 5–20 agencies).
  • More tailored workflows (rates notices, outpatient follow-up, student support) drive day-to-day use.
• Effect on adoption concentration:
  • Stronger reduction across regions covered by the consortium; more even usage across medium/small agencies.
  • Still depends on which regions form/are invited into consortia.
• Best use: place-based service clusters with similar tasks and moderate scale.

3. Embedded AI coaches in agencies (with access to a shared technical platform)

• Effect on per-capita work use:
  • High in participating agencies. Coaches translate tools into local workflows, run hands-on sessions, and normalise use.
  • Most impact where they work with managers to change routines (templates, scripts, KPIs).
• Effect on adoption concentration:
  • Strong local equalising where deployed, but coverage is thin if coaches are scarce.
  • Risk of new concentration around the few agencies that get coaches.
• Best use: targeted at under-served, high-need agencies, tied to clear service metrics.

4. Comparative pattern (most likely)

• Pure central platform: widest formal reach, lowest per-user support, modest per-capita work use, limited change in concentration.
• Consortia + central platform: better average use and more even uptake across a group of regional agencies; medium implementation complexity.
• Coaches + central platform (with or without consortia): highest per-capita work use where present; impact on concentration depends on how coaches are allocated.

5. Practical 3–5 year mix for reducing concentration

• Use a central platform as baseline infrastructure.
• Fund regional consortia as the main scaling mechanism (shared capability, localised workflows).
• Overlay a small, mobile coach program focused on:
  • Lowest-adoption regions and sectors (e.g., small councils, smaller health services, regional TAFEs).
  • Time-limited engagements that leave behind templates and peer champions, not permanent dependence.

This mix is more likely to raise regional per-capita work use and narrow hub–non-hub gaps than any single model alone, given realistic Australian constraints on skills, funding, and governance.