Methodology
How the AI Jobs Index Australia measures AI job creation, AI-attributed displacement, and the net effect on the Australian workforce.
Version 1.0 · Last reviewed 18 April 2026
On this page
1. Summary and first principles
The AI Jobs Index tracks two things every month: AI-related job creation and AI-attributed job displacement in Australia. The headline figure is the net of the two.
Four principles guide the work:
- Anchor to official statistics. The index is not a replacement for the Australian Bureau of Statistics (ABS) or Jobs and Skills Australia (JSA). It is a real-time overlay. Every metric references an official source where one exists.
- Separate exposure, capability, and observed use. A job being theoretically exposed to AI is not the same as AI being used in that job, and neither is the same as a job being replaced. The index distinguishes all three.
- Task-level, rolled up to occupation. Occupation-only analysis hides the fact that most jobs are a mix of high and low exposure tasks. Where the data allows, the index works at the task level.
- Publish limitations. Every dataset has blind spots. They are documented in section 9.
2. Definitions
Job advertisement
A publicly advertised role captured from a job board (Adzuna, Seek, Indeed, Jora). One vacancy can appear as multiple ads. We de-duplicate across boards on company + role + location where possible. This is what the index counts.
Job vacancy
An unfilled role captured by ABS through the Job Vacancies Survey (JVS). Measured directly from employers, not from advertisements. The AI Jobs Index does not publish a vacancy figure and should not be read as one. Where vacancy counts matter, we reference ABS JVS directly.
AI-specialist role
A role whose primary function is building, deploying, or governing AI systems. Examples: machine learning engineer, data scientist, applied research scientist, AI governance lead. Identified by title and described responsibilities.
AI-adjacent role
A role where AI is listed as a required or preferred tool but is not the primary function. Examples: a marketing manager required to use generative AI tools, a software engineer using Copilot. Tracked separately from AI-specialist roles.
AI-attributed displacement
A publicly announced Australian job loss where AI, automation, or machine learning is cited as a primary or secondary reason in company statements, regulatory filings, or credible media reporting. See section 5 for the classification framework.
Net AI Jobs Impact
The simple difference between AI-mentioning new job ads and AI-attributed announced job losses in the reporting period. The measures are not symmetrical (see section 9). The figure is reported with its components, not as a standalone number.
3. The five metrics
Each metric has a formal code, a single question it answers, a defined computation, and an official source it anchors to.
Australian AI Job Creation Index
What share of Australian job advertisements are for AI-related roles?
- Definition.
- Share of Australian online job advertisements containing AI-specialist titles or required AI skills, as a percentage of total advertisements, using a seven-day trailing average.
- Computation.
- (AI-mentioning ads in rolling 7 days) / (total ads in rolling 7 days) × 100. Reported as a single percentage and a 12-month trend. AI-specialist and AI-adjacent roles are reported separately.
- Cadence.
- Updated weekly (Sunday scrape, Monday aggregation). Reported in the monthly release.
- Anchors to.
- ABS Labour Force Survey for total employment context; JSA Internet Vacancy Index for national advertisement volume benchmark.
Australian AI Displacement Tracker
How many publicly announced Australian job losses have cited AI as a reason?
- Definition.
- Count of publicly announced Australian job losses where AI, automation, or machine learning is cited as a primary or secondary reason, sourced from company statements, ASX disclosures, and credible media reports.
- Computation.
- Monthly aggregate of approved layoff events, with company, sector, state, and attribution class (EXPLICIT / BLAMED / MIXED). Every event is reviewed by a human before publication.
- Cadence.
- Ingested continuously. Monthly total published on the second Thursday of each month.
- Anchors to.
- ASX announcements, company press releases, AFR, ABC, iTnews, Reuters, Bloomberg, Guardian, Startup Daily. Modelled on Challenger, Gray & Christmas (US).
Net AI Jobs Impact
What is the net of AI-related job creation and AI-attributed displacement in Australia?
- Definition.
- The difference between new AI-mentioning job advertisements and announced AI-attributed job losses for the reporting month. Reported with its components, never as a single headline without them.
- Computation.
- (AI-mentioning new ads in month) − (AI-attributed announced job losses in month). The two measures are not symmetrical (see Limitations). Positive values indicate more AI-related hiring than announced displacement.
- Cadence.
- Monthly, published on the second Thursday.
- Anchors to.
- AAJCI (creation side), AADT (displacement side).
AI Skills Demand Index
Which AI skills are growing fastest in Australian job advertisements?
- Definition.
- Top-growing AI skills requested in Australian job advertisements, split into AI Engineering Skills (ML, NLP, RAG, agents, MLOps) and AI Literacy Skills (ChatGPT, Copilot, prompt engineering).
- Computation.
- Mention count and 90-day growth rate for each skill in the AI Skills Taxonomy (section 7). Reported as top 10 growing and top 10 by volume.
- Cadence.
- Refreshed weekly. Trend reported quarterly.
- Anchors to.
- Taxonomy and method ported from LinkedIn Economic Graph / OECD.AI.
AI Exposure by Industry and Region
Which Australian industries and regions are most exposed to AI, and where is AI-related hiring concentrated?
- Definition.
- A heatmap of AI exposure by ANZSCO industry and by state / capital city, combining JSA-published automation and augmentation scores with current AI job ad volume.
- Computation.
- JSA-published exposure scores mapped to ANZSCO industries, weighted by current AI job advertisement volume per region. Reported as a quarterly heatmap.
- Cadence.
- Quarterly.
- Anchors to.
- Jobs and Skills Australia Generative AI Capacity Study for exposure scores. ABS state employment shares for weighting.
4. Data sources
Official Australian statistics (anchor sources)
- ABS Labour Force Survey . Authoritative Australian employment count. Referenced for context and denominators.
- ABS Job Vacancies Survey . Vacancies measured from employers. Used as a benchmark for total vacancy context, never mixed with ad counts.
- JSA Internet Vacancy Index . Total online advertisement volume benchmark for the AAJCI share denominator.
- JSA Generative AI Capacity Study . Automation and augmentation exposure scores by ANZSCO occupation.
Job creation data (AAJCI, AISDI)
Adzuna API (free developer tier), Seek (via Apify scraper), Indeed (via HasData), Jora (via Apify scraper). All targeting Australian AI-related job advertisements. Scrapes run weekly. De-duplication on (company + role + location) where possible.
Displacement data (AADT)
ASX announcements, company press releases, and Australian media: AFR, ABC, iTnews, Reuters, Bloomberg, Guardian, Startup Daily. Google Alerts and targeted RSS feeds capture candidate events. Every event is classified by AI (Claude) and reviewed by a human before publication. Classification framework in section 5.
Salary data
Extracted from job board structured fields and parsed from advertisement text. Supplemented with published Australian salary guides (Aquent, Morgan McKinley, Hays, Robert Half). Stored and reported in AUD ranges.
Survey data
Tech Leader Pulse: quarterly 13-question survey of Australian technology leaders distributed via email and LinkedIn. This is a convenience sample, not a probability sample. See limitations.
5. Displacement classification
Each AADT event is assigned one of three attribution classes. This is how the dashboard communicates evidence strength.
Company directly cited AI, automation, or machine learning in an official statement, press release, or regulatory filing.
A credible media outlet (AFR, ABC, Reuters, Bloomberg, Guardian, iTnews, Startup Daily) identified AI as the primary driver of the job losses, even if the company did not state this directly.
AI was cited as one of several contributing factors alongside restructuring, market conditions, or other causes.
Attribution framework adapted from jobloss.ai, with the monthly cadence and public-statement-first sourcing modelled on Challenger, Gray & Christmas.
6. Job enrichment
Each AI job advertisement captured is enriched with:
- Seniority level (entry / mid / senior / lead / director). Classified by Claude from title and description.
- Work arrangement (remote / hybrid / on-site). Extracted from structured board fields where available, classified by Claude otherwise.
- Skills extracted. Matched against the AI Skills Taxonomy (section 7). Only explicitly mentioned skills are tagged.
- Industry. Mapped from job board category data to 15 standardised Australian industry categories.
- Salary. Extracted from structured fields or parsed from description text. AUD ranges.
- Visa sponsorship. Captured from job board flags or keyword detection.
7. AI skills taxonomy
The taxonomy used for AI-specialist role identification, skills extraction, and the AISDI. Reviewed quarterly.
Languages & Tools
Frameworks
LLM & GenAI
MLOps & Infrastructure
Domains
Data Engineering
Cloud & Platforms
Governance
8. Update cadence
Monthly release: Second Thursday of each month. Headline figure is the Net AI Jobs Impact for the month prior, with AAJCI and AADT components. Timing is one day after the ABS Labour Force release so the index can be read alongside official data.
Live dashboard: Updated weekly (Sunday night scrape, Monday morning aggregation).
Displacement events: Media monitored continuously. Events classified by Claude and human-reviewed before publication.
Quarterly deep-dive: Industry, regional, and company-size cuts. Includes the AIXR heatmap refresh.
Annual State of AI Jobs in Australia: December each year. Year-in-review and 12-month trend retrospective.
Salary benchmarks: Refreshed with each quarterly deep-dive.
Pulse survey: Quarterly. Open for two weeks. Results published with the quarterly deep-dive.
9. Known limitations
Every dataset has blind spots. These are the ones that matter most.
- Creation and displacement are not symmetrical. AI job creation is measured in ads (a leading, noisy signal). AI-attributed displacement is measured in publicly announced losses (a lagging, under-counted signal). The Net AI Jobs Impact reflects that asymmetry, not a precise jobs-added-minus-lost figure. Always read the headline with both components.
- Announced ≠ actual displacement. Many companies disguise AI-driven cuts as "restructuring", "technology update", or "operational efficiency" to avoid press exposure. The AADT captures what is reported, which under-represents true AI-attributed displacement. Challenger, Gray & Christmas acknowledges the same limitation in the US.
- Scraped ads miss private hiring. Internal transfers, confidential searches, and recruiter-exclusive roles are not advertised publicly and are not captured. AI-specialist searches in particular are often conducted privately.
- One vacancy, many ads. The same role is often advertised on multiple boards. We de-duplicate on (company + role + location) where possible, but this is imperfect. Ad counts should not be read as vacancy counts. ABS Job Vacancies Survey is the authoritative vacancy figure.
- Classification is AI-assisted. Seniority, work arrangement, skills, and attribution are classified by Claude. Reviewed at the aggregate level, but individual classifications may contain errors. Displacement events are human-approved before publication.
- Salary coverage is partial. Many employers do not publish salary ranges. Coverage varies by board and role. Salary trends should be read as directional, not precise medians.
- The Pulse survey is a convenience sample. Distributed via the BTT and NTP Talent networks. Results reflect the views of engaged Australian tech leaders, not a statistically representative cross-section of the workforce.
- No forward projection. The index is a current-state and trailing-trend instrument. It does not forecast. For forecasts, refer to JSA's Employment Projections.
10. Version history
v1.0 · April 2026
- First public methodology release.
- Introduced named metrics: AAJCI, AADT, Net AI Jobs Impact.
- Anchored to ABS Labour Force Survey and Jobs and Skills Australia (JSA) as official reference sources.
- Explicit separation of job advertisements (private scrape) from job vacancies (ABS JVS).
Questions about methodology, corrections, or data partnerships: contact@aijobsindex.com.au
Want to reference the index? Cite as: "AI Jobs Index Australia, v1.0, 18 April 2026."
AI hiring in Australia, in numbers.
AI Jobs Index is an independent research project by Building Tech Teams, supported by NTP Talent.
Methodology · Data · Press · Contact james@jamesmacdonald.me