Submission 272
University of Technology Sydney (UTS)
Submitter group: Universities.
An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.
Summary
Read the original submission for the full text.
A Google Foundation-supported UTS report argues that industrial AI productivity depends on complementary skills, management, institutions and places. It proposes a national ecosystem network, durable translation institutes, assistive workforce programs, patient capital and reserved compute. Findings are correlational; the report does not establish that assistive adoption outperforms agentic adoption generally or compare place-based policy against horizontal alternatives.
Key points
Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.
Argues that AI's productivity impact will be limited until it is combined with complementary organisational, managerial and institutional capabilities.
This report recognises that artificial intelligence (AI) is a potentially transformative general-purpose technology, but one whose productivity impact will be limited until it is combined with complementary organisational, managerial and institutional capabilities
Argues that places hold the complements translating AI capability into living standards, and that these places can be built.
Above all, what is needed is a deep policy recognition and commitment that productivity is the mechanism through which AI capability translates into rising living standards, and that the drivers and complements which matter most are held by people and places. Places make the difference, and the central message of this report is that places can be made.
Says field evidence suggests industrial AI stalls between pilots and production, with management capability carrying firms across.
On the field evidence, industrial AI adoption stalls between successful pilots and production-scale use, and what carries a firm across is management capability rather than capital or technology.
Argues that Australia's research-to-innovation gap reflects industrial structure more than research quality, and that additional public research funding alone will not close it.
Australia’s research-to-innovation gap reflects industrial structure more than research quality or institutional failure, and additional public research funding alone will not close it.
Recommends that data centre approvals carry negotiated conditions reserving compute capacity for Australian research and industry.
Data centre approvals should carry negotiated conditions that reserve compute capacity for Australian research and industry, developed over time into a combined access scheme.
Says assistive AI improves on working without AI, but evidence does not yet establish superiority over agentic adoption across tasks and industries.
There is a broad understanding that assistive adoption clearly improves on working without AI, though the evidence does not yet establish that it beats agentic adoption across tasks and industries.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Make industrial AI central to the National AI Plan and NRSC terms of reference, prioritising industry capabilities across government alongside safety and adoption.
Industrial AI should be a central focus of the National AI Plan, with a role in the terms of reference for the National Resilience and Science Council (NRSC) to ensure that the development of AI capabilities in industry is prioritised across government, alongside safety and adoption initiatives.
Identify and target areas of potential competitive advantage in global markets and value chains, applying smart specialisation methods where Future Made in Australia capability, demand and ecosystem strength converge.
The strategy should identify and target areas of potential competitive advantage in global markets and value chains, applying smart specialisation methods to concentrate effort where Future Made in Australia capability, demand, and ecosystem strength converge.
Attach favourable compute reservations to data centre approvals and support for industrial AI, research institutions and SMEs; develop a combined scheme of reservations, demand-side vouchers and specialised tools.
Approvals and support for data centre investments in Australia should include negotiated conditions that reserve compute capacity on favourable terms for industrial AI applications, research institutions, and small- and medium-sized enterprises. Over time, these conditions should be developed into a combined access scheme that incorporates reservations, demand-side vouchers, and access to specialised AI tools.
Support cost-effective domestic small language models and the fine-tuning of open-weight models to Australian needs, alongside compute access.
Alongside access, government should also support cost-effective domestic small language models and the fine-tuning of open- weight models to Australian needs.
Pair compute reservations with assurance obligations covering provenance vetting of models and datasets, incident response arrangements and Australian data residency.
Negotiated access conditions should pair capacity reservations with assurance obligations, including provenance vetting of models and datasets, incident response arrangements, and Australian data residency.
Deliver reserved compute, vouchers or tools through governed local platforms with public co-funding and education partners, coupling access with local capability.
An Australian approach could be designed with reserved compute, vouchers or tool access delivered through governed local platforms involving public co-funding and education partners, so that reserved capacity is accompanied by the local capability to absorb it.
Focus R&D engagement and the application of industrial AI on place-based innovation ecosystems, developed as a national network of collaborative, industry-led nodes built on the existing precinct portfolio.
R&D engagement and the application of industrial AI should focus on place- based innovation ecosystems, developed as a national network of collaborative, industry-led nodes built upon the existing precinct portfolio.
Give each node a local specialisation, a translation-institute relationship and shared demonstration facilities, under industry-led boards with stable long-term funding.
Each node would focus on local strengths, host or partner with a research translation institute, and connect to a shared demonstration facility, all overseen by industry-led boards with stable, long term funding
Build measurement into network design: recorded baselines and theories of change for each node, matched comparator precincts rather than city averages, and longitudinal evaluation against seven indicators.
The network’s design should build in a measurement capability from the outset, with a recorded baseline and theory of change for each node, matched comparator precincts rather than city averages, and longitudinal evaluation aligned to the seven indicators
Consider statistical-system construction of precinct-scale measures from linked BLADE and PLIDA administrative data, rather than leaving measurement to individual precinct evaluations.
The statistical system could construct precinct-scale measurement from linked administrative data through the BLADE and PLIDA assets, rather than leaving it to individual precinct evaluations
Universities and CSIRO to commit strongly to industrial AI in teaching, research and engagement, including curriculums, mid-career skills training, industry-aligned research and translation institute partnerships focused on mid-sized firms.
Universities and CSIRO should make a strong commitment to industrial AI in teaching, research, and engagement. This includes developing curriculums and mid-career skills training, conducting research aligned with industry needs, and establishing translation institute partnerships specifically focused on supporting mid-sized firms.
Focus translation-institute partnerships on high-potential SMEs in collaborative ecosystems.
translation institute partnerships focused on high potential SMEs in ABOUT THE PROJECT collaborative ecosystems.
Embed industrial AI in workforce and management capability building, especially for SMEs, through extension-style services delivered by translation institutions and program design founded on assistive adoption.
Industrial AI should be embedded in workforce and management capability building, especially for small and medium- sized enterprises, through extension- style services delivered by translation institutions and program design founded on assistive adoption.
Deliver services through translation institutions, with program design based on assistive adoption and promotion of local content.
services delivered by translation institutions and program design with Dr John H Howard as Project founded on assistive adoption and promotion of local content.
Rebuild the entry-level pipeline in workforce programs through deliberate early-career development as AI thins junior roles.
Workforce programs should also rebuild the entry-level pipeline. As AI thins the junior roles through which people once built judgement and management capability, deliberate early-career development becomes part of the capability that carries firms from pilot to production
Mobilise public/private capital and public demand to finance industrial AI ventures, scale firms and secure first major customers, using NRF investment, co-investment vehicles and instruments attracting institutional capital.
Public and private capital and public demand should be mobilised together to finance industrial AI ventures, scale existing firms, and secure their first major customers. Capital should draw on the National Reconstruction Fund's investment mandate, co-investment vehicles, and instruments that attract institutional capital,
Open procurement and offtake pathways in energy, defence, health, water and major public projects to young and mid-sized firms, so public spending builds ecosystem capability as well as acquiring goods and services.
while procurement and offtake pathways in energy, defence, health, water and major public projects open to young and mid-sized firms, so that public spending builds ecosystem capability as well as acquiring goods and services.
Use NRF patient capital for firms willing to build domestic research capacity.
Increasing the number of such firms may be the intervention with the largest effect, with NRF patient capital directed towards firms willing to build research capacity domestically
Ensure public procurement and offtake pathways reward local capability and fund absorption as well as acquisition.
Offtake commitments in the energy transition, defence equipment programs, and major projects in transport, health and water create the recurring, complex challenges from which translation institutions and capable firms grow, provided the pathways reward local capability and fund absorption as well as acquisition
Use international anchor-firm attraction only with commitments to local supply chains, research partnerships and skills development.
International anchor attraction. This approach can also function as an alternative option, but only if it is supported by clear commitments. The case base includes ecosystems formed around attracted firms, whose responsibilities extend to local supply chains, research partnerships, and skills development
Frame compute reservations around inference, measured in tokens as well as racks, reserving training capacity only insofar as it is located in Australia.
A reservation scheme would therefore be most useful if framed around inference capacity, measured in tokens as much as in racks, with training capacity reserved only to the extent that it is located here.
Make voucher eligibility, allocation and claims mechanisms efficient and transparent; consider access to NISQ quantum computers.
It would, however, require mechanisms to determine eligibility, allocate entitlements and manage claims, and these would need to operate efficiently and transparently. Access to NISQ (Noisy intermediate- scale) quantum computers might also be considered in this context.
Specify clear criteria for specialised AI-tool access to prevent incumbent capture.
This scheme would need clearly specified access criteria; without them, it could drift from its purpose and fall under the control of incumbents.
Invest in translation institutes on horizons of a decade or more.
Invest in translation institutes on horizons of a decade or more, particularly as translation depends on relationships that grant cycles easily destroy.
Staff translation institutes with people respected in research and industry, mandate SME engagement and connect institutes to facilities for production-like trials.
Staff them with people who hold standing in both research and industry. ¢ Give them an explicit mandate for SME engagement. ¢ Connect them to shared facilities where adoption can be trialled in production-like conditions.
Require annual public-AI investment returns covering funding instruments, industry/use, deployment stage, skills/process/data/management expenditure, created/changed/displaced roles and the entity's success measure.
This could be met in part by a light reporting requirement attached to all public investment in AI capability, adoption programs and precinct infra- structure, collected consistently over a period long enough to observe an effect. In an annual short return, this should cover: 1. The investment and the instrument through which it was made. 2. The industry applied segment and how technology is put to use. 3. The stage reached, whether trial, pilot or production. 4. Complementary expenditure on skills, process redesign, data preparation and management capability. 5. The roles created, changed or displaced within the reporting entity. 6. The measure the entity itself uses to judge whether the investment succeeded.
Maintain reporting for at least five years on stable definitions and publish aggregate data for firms, researchers and governments.
The series must run for at least five years on a stable definition, and the data should be published in aggregate rather than held for program administration alone, so that firms, researchers and governments can all read the same record.
Give ecosystem nodes industry-led boards with real strategic authority and durable orchestrating bodies staffed by professional facilitators.
Each node needs an industry- led board with genuine authority over strategy, and an orchestrating body with long tenure, staffed as professional facilitators in the Espoo manner.
Keep national coordination light: set delivery/assessment frameworks, broker learning and aggregate demand while leaving direction to ecosystem participants.
National coordination should stay light: setting the delivery and assessment framework, brokering learning across ecosystems, and aggregating demand, while leaving direction to the ecosystem participants themselves.
Run funding on decadal horizons, blending Commonwealth, State and industry contributions, with continuation tied to performance against the seven-indicator framework rather than political cycles.
Funding should run on decadal horizons, blending Commonwealth, State, and industry contributions, with continuation tied to performance against the seven- indicator framework rather than to political cycles.
Provide translation institutes at least a decade of assured base funding before testing earnings, with review points rather than termination dates.
Build translation institute capability: durable | At least a decade of assured base funding before institutions, staffed across research and earnings expectations are tested, with review points thin industry, with a mandate for mid-sized firms rather than termination dates; the comparator institutions took ten years or more to reach their operating ratios
Make precinct commitments over ten to twenty years, staged by indicator progress rather than budget cycles.
Place-based district design: apply the four-domain and seven-indicator disciplines electoral cycles, since every district in the visits across the national precinct portfolio Ten-to-twenty year precinct commitments spanning program took at least two decades to converge; staged against measured progress on the indicators rather than the budget cycle
Commit patient capital and offtakes through first industrial-scale revenue, typically eight to fifteen years for deep technology.
Patient capital and demand: use the NRF, public procurement, and offtakes to grow research-intensive firms, funding absorption years for deep technology, beyond the reach of the Capital and offtake commitments running to first revenue at industrial scale, typically eight to fifteen forward estimates and of conventional venture fund durations
Explore deeper industrial AI collaboration through Horizon Europe calls, translation-institute partnerships, industrial data spaces and reciprocal testbed access.
The open questions are therefore about use rather than entry: targeted participation in manufacturing and AI calls, institutional partnerships with Fraunhofer, VTT and DFKI, engagement with the emerging industrial data spaces, and reciprocal testbed access for Australian firms.
Complete targeting analysis before portfolio submissions so funding follows selected areas of advantage.
Complete the targeting analysis in time to influence the portfolio submissions, so that of 2026 funding follows the selected areas of advantage rather than coming before them
Attach compute reservations to individual foreign investment approvals immediately, ahead of any general regime.
Attach compute reservation conditions to individual foreign investment approv-als from the outset, ahead of any general regime
Issue CSIRO's statement of expectations and initiate compact discussions with universities under ministerial approval.
Issue CSIRO’s statement of expectations and initiate compact discussions with universities under ministerial approval
Redirect adoption program guidelines towards extension services.
Redirect adoption program guidelines towards extension services
Approve the revised National Reconstruction Fund Investment Mandate through the Federal Executive Council as a legislative instrument, rather than through Parliament.
Approve the revised NRF Investment Mandate through the Federal Executive Council as a legislative instrument, rather than through Parliament
Negotiate a framework agreement with the states on compute reservation, since planning approvals and energy connections fall under state jurisdiction.
Negotiate a framework agreement with the states on compute reservation, since planning approvals and energy connections fall under state jurisdiction and an agreement carries more weight than deal-by-deal negotiation
Secure ministerial council engagement to open the intergovernmental pathway for the federated ecosystem network.
Secure ministerial council engagement to open the intergovernmental pathway for the ecosystem network, which operates as a federated system with multi-year cycles under state jurisdiction
Finalise the portfolio submission and offsets for translation-institute funding.
Finalise the portfolio submission and offsets for translation institute funding
Finalise the portfolio submission and offsets for capability-program scale-up.
Finalise the portfolio submission and offsets for the capability program scale-up
Finalise the portfolio submission and offsets for co-investment vehicles.
Finalise the portfolio submission and offsets for the co-investment vehicles
Begin targeted analysis under NRSC and departmental direction.
Begin targeted analysis under the direction of the NRSC and relevant depart-ments
Commission the Australian field program administratively, beginning engagement and fieldwork in 2026.
Australian field program commissioned administratively Federation calendars; multi- year cycles; engagement and field program to begin in 2026
Use the Australian field program established at launch as the validated baseline for the ecosystem network and targeting exercise.
Draw on the Australian field program, established at launch, as the validated base-line for both the network and the targeting exercise
Codify the compute access regime.
Codify the compute access regime
Establish durable funding mechanisms for translation institutes and the ecosystem network.
Establish durable funding mechanisms for the translation institutes and the eco-system network
Create a special account or statutory mandate if eventually necessary.
Create a special account or statutory mandate, if eventually necessary
Establish an Office of AI implementation unit within PM&C, reporting to the Industry and Innovation Minister and collaborating with the NRSC.
A standing governance mechanism would keep the seven components aligned. It would have three elements: ¢ An implementation unit within the Office of AI in the Department of the Prime Minister and Cabinet, reporting to the Minister for Industry and Innovation and collaborating with the National Resilience and Science Council
Publish an implementation schedule with dates for each measure and present an annual statement to Parliament.
¢ A published implementation schedule with dates for each measure ¢ An annual statement presented to Parliament.
Initially cap the compute voucher stream in the tens of millions annually, observing demand before deciding whether to scale.
an initial cap in the tens of millions a year would allow demand to be observed before any scaling decision, consistent with Section 6.5.
Read the original
In some quotes a program corrected obvious text-recognition errors for display.