Workforce, AI · 27 May 2026 · 17 min read

May 2026. Five years from the deadline we set ourselves.
InnMotion on the workforce we said we'd build. A non-commercial reform proposal · By InnMotion

We set the target. We're going backwards.

The reform plan for Australia's tech and AI workforce shortage. Six things. Plain workforce. Built around the numbers we already have and the levers we already control.

The Problem

We named the target. The workforce is moving the other way.

Australia set a national goal of 1.2 million tech workers by 2030. The Australian Computer Society revised the figure upward to 1.3 million. Three and a half years later, the workforce is shrinking.

The Department of Industry, Science and Resources reported in late 2025 that Australia had fallen to around 950,000 tech workers as of May 2025. 31,000 tech jobs were lost in the prior year. That is a 3.7% drop in a single year while the broader Australian labour market grew 2%. DISR flagged the slide as "a decline in three successive quarters throughout the year" that "indicates a potential ongoing trend." Three successive quarters is not a wobble. It is a direction.

The plan was a quarter of a million net new tech workers between 2023 and 2030. The realised result through May 2025 is a net loss. To still hit 1.3 million by 2030, Australia now needs to add roughly 70,000 net new tech workers every year for five years. The current trajectory adds zero and removes a city's worth.

1.3M

Tech workers Australia said it would have by 2030. (ACS Digital Pulse 2024.)

950K

Tech workers Australia actually has, as of May 2025. (DISR.)

200K

AI-specific jobs Australia needs by 2030. A ~500% growth in AI roles. (Tech Council of Australia, 2024.)

$16B

Annual cost of the digital skills gap to Australian business by 2030, if the gap holds. (Deloitte Access Economics / ACS.)

The AI-specific gap is the most acute.

Within that broader shortfall sits a sharper problem. The Tech Council of Australia, working with Microsoft, LinkedIn and Workday, projected in 2024 that Australia would need 200,000 AI-related roles by 2030, a roughly 500% increase in AI jobs in seven years. ACS Digital Pulse 2024 put the demand in adjacent terms: 129,000 workers skilled in AI, 134,000 in advanced data analytics, 48,000 in cybersecurity by 2030. CSIRO's 2019 AI Roadmap had already flagged the need for 161,000 new AI specialists by 2030.

Three credible bodies. Three different scopes. One consistent direction. Australia's demand for AI talent is growing in step-changes. Its supply is not. The Australian Information Industry Association reported in its 2024 Digital State of the Nation survey that 56% of Australian businesses said local AI expertise was lacking, up from 21% the year before. The perceived gap nearly tripled in twelve months.

The pipeline is contracting at the same moment demand is exploding.

The structural data is the part that should worry the Cabinet.

University ICT enrolments fell sharply this year. The Australasian Conference of Tertiary Admissions Centres reported in January 2026 that only 7,686 new undergraduate students accepted ICT-related degree offers, down from roughly 9,750 in 2025. A 21% decline year-on-year. ICT now sits at 2.9% of all undergraduate offers. The drivers cited by ACS and ACTAC are revealing: fear that AI will automate entry-level coding jobs, layoffs at the global tech giants, cost-of-living pressure pulling school leavers into immediate work, and a reorientation toward "perceived stability over tech innovation."

Graduate flow is too thin to close the gap. Australia produces roughly 7,000 IT graduates per year. To hit the 1.3 million target, ACS estimates Australia would need to add net around 60,000 tech workers per year through 2030. The graduate pipeline supplies barely a tenth.

Apprenticeship commencements are falling too. Vocational tech apprenticeship and traineeship commencements declined roughly 30% between 2023 and 2024, falling to around 151,450 across all sectors. The Free TAFE program has logged 65,000+ tech enrolments since January 2023, which is helpful, and is also dwarfed by the scale of the gap.

The gender pipeline is structurally narrow. The Department of Industry's STEM Equity Monitor shows women make up only 22% of IT enrolments and completions, 20% of engineering, and 15% of all people working in STEM jobs. In Year 12, girls represent roughly 25% of IT, physics and engineering subject enrolments. The bifurcation is locked in long before the university application form.

The research base is being defunded while we forecast demand for it. CSIRO's Data61, the federal AI and digital research unit that wrote the 2019 AI Roadmap, lost roughly 20% of its researcher headcount in the year to mid-2025, with both Data61 head Dr Jon Whittle and manufacturing head Dr Marcus Zipper announcing departures on 14 November 2025. CSIRO overall is down 12.7% of headcount, around 818 roles since July 2024, with 300 to 350 more cuts queued for 2026. The agency that named the 161,000 AI specialist target is being asked to do the work with a fifth fewer people.

The cost of doing nothing is bigger than the cost of fixing it.

RMIT Online and Deloitte Access Economics estimated in 2023 that Australia loses $9 million per day in business productivity to the digital skills gap. ACS and Deloitte put the annual cost to large business at $3.1 billion today, growing to $16 billion by 2030 if the trajectory holds. The 2026 RMIT and Deloitte update on AI literacy specifically added $18.9 billion in foregone economic growth to the bill.

The upside side of the same ledger is what makes the inaction expensive. The Productivity Commission's December 2025 report Harnessing data and digital technology modelled AI adoption as a potential 4.3% labour productivity uplift over the next decade, worth around $116 billion in GDP in the market sector, with another $19 billion annually by 2030 in the public sector. The Australian Academy of Technological Sciences and Engineering, working with OpenAI, sized the AI opportunity at up to $235 billion (6 to 8% of GDP) with $5 billion of additional government spend over five years. Different methodologies. Same direction. The number Australia leaves on the table by missing the workforce target is in the hundreds of billions.

Migration is the fastest lever and the one Australia has just reformed. The Subclass 482 Temporary Skill Shortage visa was replaced by the Skills in Demand (SID) visa on 7 December 2025, with three streams and a Specialist Skills stream offering 7-day priority processing for applicants earning above $141,210 per year. The Global Talent visa was replaced by the National Innovation Visa (Subclass 858). Both reforms are real. Both are too new to have moved the workforce numbers, and the salary threshold of the Specialist Skills stream excludes most mid-career AI engineers, who earn between $110,000 and $140,000.

The question is not whether Australia needs to do something. It is what specifically to do, in what order, at what scale, and in what year.

The Plan

Six things. Specific. Costed. Readable in one sitting.

Each reform names a number, a mechanism, and a source. None of them are new ideas in the abstract. All of them are unusually specific about scale.

A dedicated AI Talent Visa. Priority processing above the median, not above the ceiling.

Create an AI Talent Visa with 7-day priority processing for any applicant on a fixed AI / ML / data engineering occupation list earning above the Temporary Skilled Migration Income Threshold. Cap initial intake at 5,000 per year. Review annually against the 200,000-by-2030 demand model.

Why

The new Skills in Demand visa's Specialist Skills stream gives 7-day processing only above $141,210 per year. That captures directors of AI (Australian average $236,000) and senior AI architects. It does not capture the bulk of the AI workforce shortage, which sits between $110,000 and $140,000. That band is mid-career ML engineers, data engineers, applied scientists, MLOps specialists. ACS Digital Pulse 2024 projects demand for 129,000 AI-skilled workers and 134,000 advanced data analytics workers by 2030. The vast majority of those roles do not pay $141k.

Canada's Global Talent Stream gives 2-week (80th percentile) work permit processing and 10-business-day LMIA processing on a fixed occupation list with no salary ceiling above the standard wage floor. The result is the talent pipeline Australia's tech sector currently watches Canada win. Match the mechanism. Use Canada's Category B occupation list as the starter, adapted to Australian ANZSCO codes.

What stays

The Specialist Skills 7-day track for $141k+ stays. The Core Skills stream stays. The Labour Agreement stream stays. The change is purely additive: a faster track for the band of AI talent the existing reform missed.

A $24,000 reskilling wage subsidy for near-tech workers.

Pay employers a $24,000 wage subsidy for every "near tech" hire on a 12-month structured reskilling track (work plus accredited microcredential or Cert IV/Diploma). Cap at 20,000 placements per year for three years. Target 60,000 conversions by 2030.

Why

ACS Digital Pulse 2024 identifies 1.1 million Australian workers in roles adjacent to tech (finance analytics, marketing analytics, mathematics, engineering-adjacent) who could plausibly transition. The blocker is not interest. It is bandwidth: 38% of those workers cited time and financial pressure as the reason they hadn't moved. The Microcredentials Pilot is good and small. Its commitment is $18.5 million for up to 4,000 students. That is two-tenths of one percent of the near-tech pool.

The mechanism is a known one. JobMaker, JobKeeper, the Boosting Apprenticeship Commencements scheme all moved hundreds of thousands of workers via wage subsidy at scale. The Future Skills Organisation's 2025 Workforce Plan sized the shortfall at around 131,000 tech-specific workers by 2030. 60,000 funded conversions over three years closes nearly half of that, while building a more diverse tech workforce than the graduate pipeline alone produces.

What stays

The Microcredentials Pilot stays. Free TAFE stays. University funding stays. This is a third lane sitting alongside them, designed for the mid-career worker the current programs underserve.

Set a floor under CSIRO Data61 and stop cutting the agency that named the gap.

Reverse the Data61 headcount cuts in the 2026 to 2027 federal budget. Set a minimum floor of 600 FTE AI and data researchers at Data61, indexed annually to AI workforce demand. Pair with a structured industry secondment program for AI researchers.

Why

CSIRO Data61 wrote the report that told Australia it would need 161,000 new AI specialists by 2030. The agency is now being cut by 20% in a single year. The country forecasting the workforce problem is being asked to do it with fewer staff while the problem grows. The economic logic does not survive contact with the budget.

The fix is not just to reverse the cuts. It is to recognise that public AI research capability is a force multiplier on industry capability, and that the United Kingdom has just modelled this with its AI Security Institute (rebranded February 2025), which retains technical staff inside government and seconds researchers in from frontier labs. Australia has the institution. It is unwinding it. Stop unwinding it.

What stays

CSIRO's existing research mission stays. The change is a floor, not a ceiling.

A "Build Local" rule on every federal AI procurement above $5 million.

Any federal AI procurement above $5 million must include a workforce capability component: either a minimum 30% of the build delivered by Australian-resident engineers, or an equivalent investment in funded training places at an Australian university, TAFE, or industry-accredited provider. Score it in tender evaluation. Report it annually.

Why

The National AI Plan announced 2 December 2025 consolidates $460 million in existing federal AI funding plus a new AI Accelerator CRC round, alongside forecast data centre commitments exceeding $100 billion. The plan is strong on infrastructure. It is quieter on the question of who builds the AI being procured with that money. The Productivity Commission's December 2025 report returns to "workforce capability" as the binding constraint on the productivity dividend it models.

The lever exists. Australian Defence procurement already uses Australian Industry Capability scoring on sovereign capability builds. Apply the same logic to AI. The 30% local-build threshold is deliberately modest. It does not lock out global vendors. It just costs them a workforce contribution to win the contract. The alternative is the $460 million flowing predominantly offshore while domestic capability stagnates.

What stays

Open procurement stays. Best-value selection stays. Global vendors stay eligible. They just compete on workforce capability alongside technical capability.

Mandate a Year 9 AI and computing subject in every Australian school by 2028.

One semester of AI and computing in Year 9, mandatory across every state and territory by 2028. Include a half-day industry placement: every Year 9 student visits a partner organisation to see AI used in work. Track gender split as the primary KPI. Target 40% female participation by Year 12 within five years.

Why

The bifurcation that produces a 22% female tech workforce locks in around Year 9. Department of Education data on STEM subject selection shows the gender split widens dramatically between Year 9 and Year 11. By Year 12 only ~25% of IT, physics and engineering enrolments are female, and university CS / AI enrolments inherit that pool. You cannot fix the top of the funnel without fixing the middle of the school years.

A Year 9 mandatory subject is not novel. Most Australian states already mandate a digital technologies strand in Years 7 to 8 under the Australian Curriculum. The change is to extend it through Year 9 with an explicit AI focus and a workplace touchpoint, co-funded with industry under the National AI Plan. The half-day placement is the lever that converts curiosity into possibility. Year 9 students who have seen AI being used in a real workplace are measurably more likely to pick Year 11 IT or maths than students who have only studied it on paper.

What stays

State curriculum authority stays. The Australian Curriculum framework stays. The change is a minimum content requirement layered on top, not a replacement.

One National AI Workforce Authority. End the fragmentation.

Establish a National AI Workforce Authority as a statutory body reporting to the Minister for Industry, Science and Resources. Mandate: deliver a single annual AI Workforce Plan with binding intake targets across migration, university, TAFE, and reskilling. Statutory reporting to Parliament. Funded at $80 million per year for five years.

Why

Tech workforce policy is currently split across at least eight federal bodies: the Department of Industry, Science and Resources (industry strategy), the Department of Employment and Workplace Relations (skills), the Department of Education (universities), the Department of Home Affairs (migration), Jobs and Skills Australia (forecasting), the Future Skills Organisation (workforce planning for finance / tech / business), the Digital Transformation Agency (public service workforce), and CSIRO (research). Each produces credible reports. No single body has end-to-end accountability.

Singapore solved exactly this with the Infocomm Media Development Authority. IMDA owns the TechSkills Accelerator (TeSA), the National AI Impact Program, the digital industry strategy, and the workforce reporting line. The output: 340,000 Singaporeans upskilled in tech since 2016, with another 40,000 tech professionals committed to AI upskilling over three years from late 2025. The instrument exists. Australia has eight smaller instruments where Singapore has one. The cost of consolidating is real. The cost of continuing the fragmentation, on Deloitte's numbers, runs to $16 billion a year by 2030.

What stays

JSA continues to do labour market forecasting. The Future Skills Organisation continues to plan finance and business workforce. Universities and TAFEs continue to deliver. The Authority is the integrator and the accountable body, not a replacement for delivery.

Bonus. Publish the workforce numbers monthly. Not annually.

A monthly Tech and AI Workforce Report: total tech workers (direct and indirect), AI-specific role count, university enrolments year-on-year, TAFE commencements, migration intake on tech occupations, and gap-to-target. Same cadence as the monthly Labour Force Survey.

Why

The reason the country only noticed the workforce was shrinking in late 2025 is that the quarterly DISR series is the only public series, and even it lags. The Australian Bureau of Statistics publishes labour force data monthly. The tech and AI workforce, currently the central productivity question for Australia, gets quarterly attention. Make it monthly. Daylight builds urgency. Urgency builds policy response. Quarterly reporting builds press releases.

International Lessons

What the countries Australia compares itself to are doing.

Three of them are doing more, faster. One is doing the same fragmented thing Australia is doing. The fifth has solved the talent-attraction half of the problem and is now exporting the playbook.

Singapore. The integrator model.

The IMDA runs the TechSkills Accelerator (TeSA), which has upskilled over 340,000 individuals in tech skills since 2016. Under the National AI Impact Program, IMDA committed in late 2025 to upskilling 40,000 tech professionals in AI over three years, including final-year Information and Digital Technologies students. The program covers AI, cybersecurity, data analytics, software engineering, network and infrastructure, and generative AI. Salary support grants run up to 12 months for trainees.

The lesson is structural, not programmatic. Singapore put one statutory authority in charge of every lever: industry policy, skills development, AI strategy, and workforce reporting. The result is a single accountable line, a single integrated plan, and a single number to report against. Australia has eight bodies doing parts of the same job. The integrator model is the most copyable thing here.

United Kingdom. The AI Opportunities Action Plan.

The UK published its AI Opportunities Action Plan on 13 January 2025, a 50-point program targeting £400 billion of additional GDP by 2030. The talent-attraction levers were the most ambitious: expansion of the High Potential Individual visa (graduates of around 40 top global universities, no job offer required), a dedicated internal headhunting capability to acquire elite AI talent, and the Invest 2035 Industrial Strategy positioning AI as a priority sector. The AI Security Institute (rebranded February 2025) is funded explicitly to retain technical staff inside government.

The lesson for Australia is the combination of a high-end visa lane with public-sector AI capability. The UK is acquiring senior AI talent and using government as a buyer-of-last-resort to keep that talent in the country. Australia's National Innovation Visa is the analogue. The UK has demonstrated the playbook.

Canada. The Global Talent Stream.

Canada's Global Talent Stream delivers 2-week (80th percentile) work permit processing and 10-business-day LMIA processing for employers on a fixed occupation list (software engineers, IT analysts, electrical engineers). Category A handles specialised talent (advanced degree or 5+ years specialist experience). Category B handles roles on the published Global Talent Occupations List. There is a permanent residence pathway via Express Entry.

The lesson is that speed is a policy choice. Canada's process is fast because Canada decided it would be fast. Australia's new Skills in Demand Specialist Skills stream is now competitive on speed (7 days versus 10 business days) but at a much higher salary threshold. The mechanism to copy is Canada's: a published occupation list, a fast-track for everyone on it, and a permanent residence ladder behind it.

United States. The talent magnet, by inertia.

The US has the deepest AI talent pool by an order of magnitude, but the supply system is straining. H-1B caps have not risen materially since 1990, and the AI talent that wants to be in the US is increasingly going to Canada (because the visa is faster) and the UK (because the visa is easier). The lesson for Australia is the negative one: do not assume the US will keep absorbing the world's AI talent indefinitely. The window for Australia to attract talent that would otherwise go to Silicon Valley is widening, not narrowing.

Germany. The fragmented cautionary tale.

Germany has named comparable targets, run comparable upskilling schemes, and shipped a national AI strategy in 2018 with a €5 billion budget. Germany is also missing its targets in roughly the same way Australia is, for roughly the same reason: federal-state fragmentation, a slow migration pipeline, and a research base that is high-quality and undersized for the demand. Germany is the warning of what happens when you set ambitious AI workforce goals without consolidating accountability.

India. The exporter, not the importer.

India's National AI Mission and the IndiaAI portal are doing the inverse of every country above: building scale capability at home and exporting AI talent globally. India already supplies a meaningful share of the AI engineers hired by Australian firms via the Skilled Independent visa. The strategic question for Australia is whether to keep importing Indian AI talent role-by-role or to negotiate a structured bilateral framework, the way Germany has done with India's IT services sector. The capability is real. The relationship needs work.

The Principles

Five things we hold as non-negotiable.

Name a single number. 1.3 million by 2030. Report against it monthly. Hold the policy machinery to it.
One accountable body. Fragmentation across eight departments is the policy failure. Consolidate it.
Migration is the fastest lever. Use it like one. Match Canada on processing speed. Match the UK on top-end attraction. Set the salary floor at the median tech wage, not the ceiling.
The pipeline starts in Year 9, not Year 12. Fix the school-age gender bifurcation before it becomes the university enrolment bifurcation, before it becomes the workforce bifurcation.
Don't defund the agency that wrote the report. CSIRO Data61 forecasting a 161,000 AI specialist gap while losing 20% of its staff is the policy equivalent of self-harm.

What we don't pretend to know

The open questions, honestly.

Anyone selling you a workforce plan with no holes in it is selling you something else. Here are the four we know are still open.

1. How AI itself reshapes the demand curve.

The same technology driving demand for 200,000 AI specialists is also reshaping the demand for the broader software engineering workforce. ACS reported in 2025 that Australian firms are hiring fewer entry-level engineers because AI tooling has absorbed much of that work. If that pattern holds, the 1.3 million number may overstate human-headcount demand. It may also understate AI-skilled-headcount demand. The honest answer is we don't know yet, and the plan above is robust to both directions: more migration, more reskilling, more local capability is the right move in either scenario.

2. Whether $24,000 is the right reskilling subsidy.

ACS proposed the figure in its 2025 recommendations. It is roughly half the median annual tech wage and matches the JobMaker tier-1 quantum. It is also untested at the proposed scale of 20,000 placements per year. The principle is right: scale-matched subsidy beats boutique microcredential funding. The exact number wants a year of trialling at 2,000 to 3,000 placements before the full rollout.

3. The state and territory politics of a national curriculum mandate.

Education is a state responsibility. A Commonwealth mandate on Year 9 curriculum runs into thirty years of jurisdictional friction. The mechanism is achievable. National Schools Reform Agreements, conditional funding ties, joint COAG-equivalent agreements. It is not technical. It is political. The plan above assumes a Commonwealth-led negotiation. That assumption may be optimistic.

4. Whether 5,000 dedicated AI visas a year is the right cap.

Canada issues materially more than 5,000 fast-tracked tech visas a year. The UK is open-ended on the High Potential Individual visa. Australia is starting from a smaller base of senior AI capability and a smaller domestic AI labour market. 5,000 is a plausible start. It may be the right number forever. It may be the right number for 2027 and need to double by 2030 once the domestic pipeline starts catching up. The annual review built into Reform 1 is where that question gets answered.

Build the country we said we'd build.