Three government evidence packages now say the same thing in different accents: artificial intelligence is reshaping occupational demand unevenly, and the labour market as a whole has not collapsed into upheaval. Australia’s Department of Employment and Workplace Relations reports no broad AI-driven disruption through February 2026, while occupations most exposed to generative AI grew 5.6% against 9.5% for the least exposed since November 2022. The U.S. Bureau of Labor Statistics projects strong 2024–34 growth in computer and mathematical roles alongside declines in several office, administrative, and customer-service occupations. The 2026 Economic Report of the President frames mixed employment evidence and Jevons’ Paradox as the operator question, not a binary jobs apocalypse.
The defended position is measurement before narrative. Responsible AI teams that brief boards on “AI and jobs” need occupation-level exposure monitoring, named reskilling owners, and hiring-path KPIs that track exposed roles separately from headline unemployment. Proof that employment growth in highly exposed occupations has accelerated past less-exposed peers, with youth outcomes collapsing, would undercut the claim. The primary record available in mid-2026 does not show that pattern.
What Australia measured through February 2026
On July 8, 2026, DEWR’s Office of the Chief Economist published AI and employment in Australia, a monitoring framework rather than a forecast. The department states there is no evidence to date of broad AI-driven labour-market upheaval. Aggregate conditions remain strong by historical standards, youth outcomes have mostly held up, and occupational reshuffling has not accelerated. Minister Amanda Rishworth’s release pairs that finding with May 2026 labour-market context: unemployment at 4.4% and participation near record highs.
The concentrated signal sits inside the aggregate calm. Between November 2022 and February 2026, employment in the most-exposed fifth of occupations grew 5.6%, compared with 9.5% in the least-exposed fifth. DEWR’s core model implies that an occupation one standard deviation above average AI exposure had employment about 2% lower by February 2026 than under its pre-ChatGPT trend. The report labels the finding suggestive, not definitive, and notes sensitivity to alternative exposure measures. Software and applications programmer employment rose 25% over the same window, which undercuts any claim that “tech jobs” moved as one block.
What BLS projects for U.S. occupations, 2024–34
BLS’s July 16, 2026 Economics Daily brief translates Employment Projections into an AI and IT lens. Data scientists are projected to grow 33.5% (about 82,500 jobs). Information security analysts grow 28.5%, actuaries 21.8%, operations research analysts 21.5%, and computer and information research scientists 19.7%. Software developers grow 15.8%, adding roughly 267,700 jobs, the largest absolute gain among the highlighted occupations. Total employment across all occupations is projected up 3.1%.
The other side of the table is equally specific. Customer service representatives are projected down 5.5% (about 153,700 jobs). Legal secretaries and administrative assistants fall 5.8%, procurement clerks 8.7%, claims adjusters 5.1%, and several other office and administrative titles decline as AI-driven productivity gains dampen demand. BLS frames the pattern as uneven occupational change inside a growing labour market, not as net employment disappearance.
How the 2026 Economic Report of the President frames the same fork
Chapter 5 of the 2026 Economic Report of the President, The Revolution of Artificial Intelligence, reviews mixed employment evidence. It cites work showing employment pressure for early-career workers in AI-exposed occupations such as coding and customer service, studies finding no correlation between AI exposure and current unemployment rates, and research where AI exposure raises employment in sectors reliant on AI-capable tasks even as it substitutes elsewhere. Overall U.S. unemployment is cited at 4.4% as of December 2025.
The chapter’s operator frame is Jevons’ Paradox: efficiency gains can expand total use of a resource when cost savings raise demand faster than unit labour needs fall. Historical analogies from steam, electricity, computers, and the internet point toward eventual employment and earnings gains, while the report flags agency and extreme productivity without new demand as the scenarios that would break that pattern. For Responsible AI briefs, the chapter is useful less as a verdict and more as a monitoring checklist: track early-career paths in exposed occupations, firm-level AI use, and whether productivity shows up in wages and new task creation.
Adoption is rising, and that is the operator input
Census Bureau analysis of Business Trends and Outlook Survey data from December 14, 2025 to May 3, 2026 finds overall business AI use between 17% and 20%, with 20% to 23% of businesses expecting use within six months. Larger firms lead: 37% of firms with at least 250 employees reported AI use, and 32% of firms with 100 to 249 employees did so in the period ending May 3, 2026. Information (39.7%) and Finance and Insurance (33.9%) sit above the national rate near 19.8%, while Retail Trade sits near 14%. After November 2025, BTOS asks about AI in any business function rather than only production of goods and services, so operators should not treat the series as a continuous production-only measure.
The ERP chapter’s production-use series, drawn from Census BTOS, shows firms using AI to produce goods and services rising from under 4% in 2023 to about 10% by September 2025. Rising adoption is the reason occupation-level monitoring matters now. If tools diffuse while exposed clerical and customer-service paths soften relative to computer and mathematical growth, workforce governance has to move at adoption pace, not at headline unemployment pace.
What Responsible AI operators should change now
The Factics move treats the DEWR July 2026 monitoring report, the BLS 2024–34 AI and IT projections brief, ERP Chapter 5, and the Census BTOS May 2026 business-use analysis as verified primary evidence. The tactic is a three-layer workforce checkpoint before major AI workflow expansion: (1) map roles by AI exposure using an agency or internal score, (2) assign a named owner for reskilling and redeployment in the top-exposed quintile, and (3) separate early-career hiring KPIs for exposed occupations from firmwide headcount. The KPI is quarterly: share of high-exposure roles with a documented learning path, year-over-year employment or hours change in those roles versus low-exposure peers, and early-career hire rates in computer/math versus customer-service and routine administrative paths.
Checkpoint-Based Governance supplies the authority test. A headline that “AI is taking jobs” or “AI creates jobs” does not move a workforce decision by itself. If a deployment automates customer contact, claims review, or administrative drafting, the checkpoint question is who owns the exposure map, who funds the reskilling path, and who reports occupation-level KPIs to leadership. Factics keeps the brief honest: fact, tactic, and KPI travel together, or future-of-work governance collapses into narrative. HAIA-CORE is the evaluation method that forces those claims into readable structure for operators who brief boards on the same evidence.
Watch the next data releases rather than the loudest forecast. DEWR plans ongoing monitoring under the same framework. BLS will refresh occupational projections. Census will keep publishing BTOS AI supplements with function-level detail. Until those series show broad upheaval, treat uneven occupational change as the operating reality and staff the checkpoint accordingly.
Sources
- U.S. Bureau of Labor Statistics. (2026, July 16). Artificial intelligence, information technology, and employment, 2024–34. The Economics Daily. https://www.bls.gov/opub/ted/2026/artificial-intelligence-information-technology-and-employment-2024-34.htm
- Department of Employment and Workplace Relations, Office of the Chief Economist. (2026, July). AI and employment in Australia: Monitoring framework and evidence to date. Australian Government. https://www.dewr.gov.au/download/17666/ai-and-employment-australia/43205/ai-and-employment-australia/pdf
- Department of Employment and Workplace Relations. (2026, July 8). The AI and employment in Australia report [Announcement]. Australian Government. https://www.dewr.gov.au/workplace-relations/announcements/ai-and-employment-australia-report
- Rishworth, A. (2026, July 8). Australia’s labour market remains resilient amid the rise of AI [Media release]. Ministers’ Media Centre, Australian Government. https://ministers.dewr.gov.au/rishworth/australias-labour-market-remains-resilient-amid-rise-ai
- Council of Economic Advisers. (2026). Economic Report of the President (Chapter 5, The revolution of artificial intelligence). The White House. https://www.whitehouse.gov/wp-content/uploads/2026/04/ERP-2026-5.-The-Revolution-of-Artificial-Intelligence.pdf
- Grundy, A., Breaux, C., & Khatiwoda, D. (2026, May 26). Large firms with at least 20 employees biggest AI users. U.S. Census Bureau. https://www.census.gov/library/stories/2026/05/ai-use-businesses.html
Frequently Asked Questions
Is AI causing broad job loss in Australia according to DEWR?
No. DEWR’s July 2026 report finds no evidence to date of broad AI-driven labour-market upheaval. Aggregate conditions remain strong, youth outcomes have mostly held up, and occupational reshuffling has not accelerated. The report does find slower growth in more AI-exposed occupations and treats that signal as justification for ongoing monitoring.
What is the 5.6% versus 9.5% employment growth finding?
Between November 2022 and February 2026, employment in Australia’s most AI-exposed fifth of occupations grew 5.6%, compared with 9.5% in the least-exposed fifth. DEWR and the minister’s release present this as gradual occupational shift rather than proof of large-scale AI job destruction.
Which U.S. occupations does BLS expect to grow or shrink with AI and IT?
For 2024–34, BLS projects strong growth for data scientists (33.5%), information security analysts (28.5%), and related computer and mathematical roles, with software developers adding the largest absolute job count among highlighted titles. Declines are projected for customer service representatives (−5.5%), several secretary and administrative titles, procurement clerks, and related office support roles.
What does the 2026 Economic Report of the President say about AI and employment?
Chapter 5 describes mixed evidence: early-career pressure in some exposed occupations, no clear link between AI exposure and overall unemployment in other studies, and cases where AI exposure raises employment in AI-reliant sectors. It uses Jevons’ Paradox to explain how productivity gains can expand labour demand when cost savings raise output demand.
How many U.S. businesses report using AI in Census BTOS data?
Census analysis of BTOS from mid-December 2025 through early May 2026 finds overall AI use between 17% and 20%, with higher rates among larger firms (37% for 250+ employees). Sector rates are highest in Information and Finance and Insurance. Wording now covers AI in any business function after a November 2025 revision.
What should Responsible AI operators change after these releases?
Map roles by AI exposure, assign a named owner for reskilling in the highest-exposure quintile, and report early-career and occupation-level KPIs separately from headline unemployment. Treat uneven growth as the operating fact until monitoring series show broad upheaval, and tie each major AI workflow expansion to those checkpoints.
#AIgenerated
Leave a Reply