Methodology & figures
How we measure which real jobs an AI agent could do, how every figure is calculated, and how we extrapolate to all advertised jobs in Austria and the OECD.
Current figures
What we measure
Analysed jobs
All job ads that have gone through the assessment since botjobs started in August 2026, counted per source portal.
Agent-capable jobs
Jobs that passed both stages of the assessment and are online right now. Expired jobs drop out of this count.
AI quota
Agent-capable jobs listed today ÷ assessed jobs that are still advertised on their source portals today. Both numbers refer to the same population: jobs advertised right now. This share is the signal botjobs measures; the extrapolation is derived from it.
Advertised annual salaries affected
Sum of the gross annual salaries stated in the listed ads.
Extrapolation
The same share applied to the officially counted open vacancies in Austria and in the OECD countries.
How to read botjobs
botjobs mixes five kinds of statement. The harder the statement, the closer its basis is shown. Each figure on this site belongs to exactly one of them.
| Level | How botjobs handles it | On this site |
|---|---|---|
| Facts | Stated hard. Must be true. | Number of listed jobs, advertised salaries, the studies we cite (Stanford, ILO, OECD). |
| AI assessments | Shown as machine verdicts and recognisable as such. | The AI suitability score of each job (0–100) and its one-sentence reasoning. |
| Extrapolations | May be dramatic; the assumption stands right next to them. | Austria and OECD: our quota applied to the official vacancy counts. |
| Scenarios | May be dystopian; never presented as something already observed. | The red “Jobs killed by AI” counter and its cost figures (non-replacement scenario). |
| Opinion and provocation | Maximum freedom: questions, contradictions, accusations. | The manifesto, “Who pays?”, “Who owns the machines?”, “After humans?”. |
Data source
All jobs come from established Austrian job portals of the candidatis network. Every job is a real vacancy of a real employer; applications go through the normal process of the source portal. Only jobs in Austria are listed. Jobs behind a login on the source portal are not listed, and paid premium ads are never touched.
| Source portal | Jobs analysed |
|---|---|
| tecjobs.at | 17,638 |
| wujobs.at | 8,483 |
| informatikjobs.at | 1,134 |
| bankkarriere.at | 890 |
| jusjobs.at | 613 |
| justjobs.at | 331 |
| pharmakarriere.at | 233 |
| Total | 29,322 |
How a job is assessed
Each job is assessed twice by a language model:
Automation score (0–100)
How much of the core work is digital and language-based, so that a frontier AI agent could largely carry it out. Fixed caps apply for physical work, empty ads and legally protected professions. Only jobs scoring 70 or more continue.
Role check
Could an agent fill the role as a whole? Roles that depend on sales relationships, leading people, liability or physical presence are excluded. If in doubt: no.
How the figures are calculated
AI quota
agent-capable jobs listed today ÷ assessed jobs still online on their source portals × 100. The denominator comes from the latest stock snapshot after each assessment run; jobs that have expired since are in neither number.
Annual salaries
Austrian job ads must state a minimum pay. We take the stated monthly gross minimum (if only a maximum is given, that), multiply it by 14 (the Austrian 13th and 14th monthly salary) and add it up over all listed jobs that state a salary. Jobs without a salary are not counted, so the sum is a lower bound.
Live counter
The home page reloads the figures every minute from /api/stats. Numbers only move when the stock of jobs actually changes; there is no simulated movement. The red counter is different: it is a model that runs continuously (see below).
Extrapolation: Austria and the OECD
Our jobs are a sample. The sample gives the signal; with the signal we estimate the impact on all advertised jobs. The method is deliberately simple and fully visible:
today's signal: agent-capable share of assessed jobs
- Signal: the current AI quota of botjobs.
- Base: the latest official count of open vacancies per country (national statistical offices and Eurostat). For Austria that is Statistik Austria.
- Agent-capable jobs = open vacancies × signal, per country.
- Annual salaries affected = agent-capable jobs × average annual wage of the country (OECD, full-time equivalent, converted to euros at market exchange rates).
- Taxes & social contributions at risk = salaries affected × (income tax + employee and employer social security contributions, incl. payroll taxes) as % of the gross wage, per country (OECD Taxing Wages, average wage, single person). This is what the public purse would lose if these jobs disappeared without replacement.
- Pension contributions = salaries affected × mandatory pension contribution rate (employee + employer) of the country (OECD Pensions at a Glance). They are part of the taxes & contributions above, shown separately because they fund today's pensions.
- Working hours = agent-capable jobs × average annual hours actually worked per worker in the country (OECD).
- OECD figure = sum over all OECD countries that publish a vacancy count.
What the extrapolation can and cannot say
- The signal comes from Austrian job ads, and the mix of our source portals is not the mix of the economy: IT and engineering jobs are over-represented. The extrapolation therefore assumes that the share holds elsewhere. It is an order-of-magnitude estimate, not a statistic.
- Vacancy counts are defined differently from country to country (survey vs. registered openings, seasonally adjusted or not, monthly or quarterly). The OECD total is an indicative sum. Japan counts only openings registered with the public employment service; Korea uses a narrow definition.
- The wage is the average wage of all employees in a country, not of the agent-capable jobs. Agent-capable jobs are mostly qualified office and tech roles and are often paid above average, so this side tends to understate.
- Tax and pension rates are those of a single person at the average wage. Taxes are progressive, so for the better-paid agent-capable jobs the true loss is likely higher. Some pension rates also fund disability benefits; in a few countries pensions cannot be separated from other social security (Belgium, Ireland, Norway, Spain, UK).
- Working hours: the OECD notes that comparing hour levels between countries is of limited reliability; Türkiye uses 2021.
- Six OECD countries publish no absolute vacancy count and are left out instead of being estimated: Italy, New Zealand, Mexico, Chile, Colombia, Costa Rica.
The "jobs killed by AI" counter
Companies rarely announce that AI replaced someone. They simply do not refill the position when it becomes vacant. That is what the red counter on the home page models, worldwide, from 1 January 2023:
- Workers worldwide: 3.6bn (ILO, Employment and Social Trends 2026: 186 million unemployed are 4.9 % of the labour force).
- × share of jobs exposed to generative AI: 25.0% (ILO/NASK, Generative AI and Jobs: A Refined Global Index of Occupational Exposure, 2025).
- × hires per job and year: 39.6% (US JOLTS, August 2026: hires rate 3.3 % a month). This is how often a position is filled anew.
- × share of employees working in companies that use AI (Eurostat, by company size, weighted by employment): 15.2 % in 2021, 16.2 % in 2023, 23.6 % in 2024, 32.9 % in 2025, continued with the last annual increase of 9.3 points – today 45.2%.
- Assumption: every employer that uses AI no longer refills an exposed position when it becomes vacant. Jobs killed = sum over time since 1 January 2023.
What one lost job costs
A job that disappears does not only mean lost income tax. The state pays unemployment benefits, the household spends less (less VAT, less turnover for others) and saves less, and the lost spending costs further output. Per job at the average wage and per year:
| Austria | OECD average | |
|---|---|---|
| Gross wage | €58,754 | €60,581 |
| 1 Income tax + social contributions lost (employee and employer) | €35,305 | €22,325 |
| 2 Unemployment benefits paid by the state (first year) | €21,422 | €12,949 |
| 3 Household income lost (net wage minus benefits) | €18,249 | €32,201 |
| of which less consumption | €16,113 | €30,355 |
| of which less saving | €2,135 | €1,846 |
| VAT and other consumption taxes lost | €3,352 | €2,543 |
| 4 Second round: output lost through lower demand (× 0.6) | €10,949 | €19,321 |
| Cost to the state (1 + 2 + consumption taxes) | €60,079 | €37,817 |
| in % of the gross wage | 102.3% | 62.4% |
| Economic output gone (labour cost + second round) | €85,925 | €86,796 |
Under our first-year fiscal loss scenario, every job lost in Austria costs the state 102.3% of what the person earned – more than the gross wage itself.
On the home page, these per-job costs are applied to the killed jobs in OECD countries only (18.5% of world employment, OECD Employment Outlook 2025: 668 million); outside the OECD there are no comparable data.
Cost of one lost job per country
| Country | Benefit % of net | Saving % | Consumption tax % | Cost to the state | % of gross |
|---|---|---|---|---|---|
| United States | 19 | 5.7 | 5.4 | €38,411 | 49.9% |
| Japan | 44.2 | 0.7 | 12.1 | €23,831 | 77.7% |
| Germany | 60 | 10.3 | 17.2 | €54,904 | 100.1% |
| United Kingdom | 11 | 4.3 | 16.1 | €30,288 | 55.8% |
| Canada | 34.8 | 4.5 | 13 | €36,387 | 67.1% |
| France | 68 | 12.8 | 20.6 | €52,032 | 117.4% |
| Netherlands | 69.8 | 10.5 | 23.8 | €58,698 | 95.5% |
| Australia | 25 | 6.1 | 11.4 | €33,032 | 54.9% |
| Türkiye | 35 | – | 18.5 | €16,811 | 80.5% |
| Korea | 37.7 | 10.2 | 12.5 | €21,797 | 64.9% |
| Spain | 52 | 9.2 | 16.2 | €34,864 | 99.3% |
| Israel | 29.5 | – | 20.9 | €33,270 | 61.7% |
| Belgium | 57.8 | 6.6 | 18.2 | €65,485 | 106.1% |
| Austria | 54 | 11.7 | 20.8 | €60,079 | 102.3% |
| Sweden | 47.3 | 13.6 | 24.4 | €48,982 | 99.2% |
| Switzerland | 72 | – | 9.6 | €89,737 | 85.5% |
| Poland | 24.5 | 6.1 | 20.8 | €16,811 | 70.4% |
| Norway | 63 | 7 | 23.1 | €57,761 | 92.1% |
| Czechia | 22.9 | 12.9 | 20.7 | €21,769 | 84.1% |
| Denmark | 52.2 | 9.8 | 27.1 | €55,992 | 77.3% |
| Hungary | 11.5 | 14.7 | 30.6 | €13,991 | 69.5% |
| Portugal | 87 | 3.6 | 20.9 | €29,947 | 116.0% |
| Greece | 35 | -9.3 | 20.1 | €15,684 | 83.4% |
| Lithuania | 50.5 | 3.8 | 20.6 | €24,742 | 77.5% |
| Ireland | 28 | 9 | 20.5 | €42,662 | 67.3% |
| Finland | 43 | 6.1 | 25.6 | €45,744 | 90.5% |
| Slovenia | 31.5 | 6.09 | 23.5 | €31,010 | 82.5% |
| Latvia | 36.3 | 0 | 22.4 | €21,484 | 87.0% |
| Slovakia | 33 | 2 | 18.2 | €19,372 | 90.5% |
| Estonia | 46.4 | 3 | 24.9 | €28,270 | 102.6% |
| Luxembourg | 86 | 5 | 32.1 | €92,058 | 107.0% |
| Iceland | 47.8 | 6.32 | 25.3 | €65,061 | 77.3% |
This is a deliberately pointed model, not a count. Its central assumption – employers using AI stop refilling exposed positions – is a thought experiment; "exposed" in the ILO sense means transformed, not necessarily replaced. US hiring rates and EU adoption rates are applied worldwide, and killed jobs are not assumed to be re-absorbed elsewhere. The output figure assumes the work is not done at all; in reality much of it would be done by AI instead, so economic output does not simply disappear – what changes is who earns it. Unemployment benefits are those of the first year; after that they fall in most countries while the lost taxes remain. Change any of these numbers and the counter changes with them.
ILO Employment and Social Trends 2026: ilo.org · ILO/NASK WP 140 (2025): ilo.org · BLS JOLTS August 2026: bls.gov · Eurostat isoc_eb_ai, sbs_sc_ovw: ec.europa.eu
OECD Benefits and wages, net replacement rate in unemployment, single, 100% of average wage, average of months 1-12, without social assistance and housing benefit (2025; JP, NL, DK 2024). OECD Household savings, net saving as % of household net disposable income (2024/2025); missing for TR, CH, IL -> OECD sample average. OECD Revenue Statistics: taxes on goods and services (5000) / household final consumption expenditure (2024; AU, GR 2023). Multiplier: Gechert (2015), Oxford Economic Papers 67(3), meta-analysis of 104 studies: https://doi.org/10.1093/oep/gpv027.
Vacancies, wages, taxes, pensions and hours per country
| Country | Open vacancies | Period | Average annual wage | Taxes & contrib. % of gross | Pension % | Hours/year | Source |
|---|---|---|---|---|---|---|---|
| United States | 7,079,000 | Aug 2026 | €76,972 | 32.4 | 10.6 | 1,800 | data |
| Japan | 2,227,740 | Aug 2026 | €30,677 | 38.31 | 18.3 | 1,598 | data |
| Germany | 1,017,924 | 2026Q2 | €54,840 | 59.54 | 18.6 | 1,332 | data |
| United Kingdom | 691,000 | Jul-Sep 2026 | €54,285 | 36.8 | 26.6 | 1,533 | data |
| Canada | 501,035 | Jul 2026 | €54,213 | 35.21 | 9.2 | 1,687 | data |
| France | 484,441 | 2026Q2 | €44,304 | 64.33 | 27.8 | 1,498 | data |
| Netherlands | 387,200 | 2026Q2 | €61,492 | 40.47 | 22 | 1,436 | data |
| Australia | 325,000 | Aug 2026 | €60,212 | 29.6 | 11.5 | 1,633 | data |
| Türkiye | 170,023 | 2026Q2 | €20,885 | 47.8 | 20 | 1,732 | data |
| Korea | 160,000 | Aug 2026 | €33,583 | 27.6 | 9 | 1,833 | data |
| Spain | 150,765 | 2026Q2 | €35,121 | 54.11 | 28.3 | 1,633 | data |
| Israel | 148,047 | Sep 2026 | €53,930 | 27.7 | 19.3 | 1,870 | data |
| Belgium | 133,881 | 2026Q2 | €61,747 | 66.77 | 16.4 | 1,586 | data |
| Austria | 123,399 | 2026Q2 | €58,754 | 60.09 | 22.8 | 1,442 | data |
| Sweden | 122,200 | 2026Q2 | €49,354 | 54.06 | 22.3 | 1,421 | data |
| Switzerland | 98,700 | 2026Q2 | €104,942 | 24.45 | 16.6 | 1,512 | data |
| Poland | 98,400 | 2026Q2 | €23,881 | 40.74 | 19.52 | 1,776 | data |
| Norway | 96,776 | 2026Q2 | €62,703 | 41.12 | 22.8 | 1,386 | data |
| Czechia | 87,217 | 2026Q2 | €25,890 | 55.13 | 31.3 | 1,824 | data |
| Denmark | 77,473 | 2026Q2 | €72,422 | 36 | 12.7 | 1,369 | data |
| Hungary | 67,434 | 2026Q2 | €20,129 | 46.5 | 21.6 | 1,671 | data |
| Portugal | 65,721 | 2026Q1 | €25,826 | 48.68 | 22.7 | 1,764 | data |
| Greece | 46,368 | 2026Q2 | €18,801 | 47.91 | 26 | 1,874 | data |
| Lithuania | 31,910 | 2026Q2 | €31,939 | 40.51 | 8.72 | 1,673 | data |
| Ireland | 30,000 | 2026Q2 | €63,379 | 36.28 | 14.1 | 1,639 | data |
| Finland | 29,338 | 2026Q2 | €50,569 | 51.17 | 24.81 | 1,492 | data |
| Slovenia | 20,739 | 2026Q2 | €37,569 | 52.81 | 24.35 | 1,599 | data |
| Latvia | 19,423 | 2026Q2 | €24,685 | 49.61 | 20 | 1,588 | data |
| Slovakia | 18,297 | 2026Q2 | €21,403 | 56.47 | 22.8 | 1,619 | data |
| Estonia | 8,306 | 2026Q2 | €27,554 | 57.05 | 20 | 1,610 | data |
| Luxembourg | 6,900 | 2026Q2 | €86,017 | 45.64 | 16 | 1,459 | data |
| Iceland | 6,891 | 2026Q2 | €84,113 | 33.52 | 21.8 | 1,421 | data |
Wages: OECD Average annual wages, per full-time equivalent dependent employee, 2024/2025, current prices, converted to euros at the ECB annual average exchange rate. Research of the base data: 25 October 2026. OECD Data Explorer
OECD Taxing Wages 2026 (figures for 2025), single person without children at 100% of the average wage: income tax + employee and employer social security contributions incl. payroll taxes, as % of gross wage. https://data-explorer.oecd.org/vis?df[ds]=dsDisseminateFinalDMZ&df[id]=DSD_TAX_WAGES_DECOMP%40DF_TW_DECOMP&df[ag]=OECD.CTP.TPS
OECD Pensions at a Glance 2025, Table 8.1 (mandatory contribution rates 2024, effective rate on average earnings, employee + employer, public + mandatory private): https://stat.link/orhufb
OECD Average annual hours actually worked per worker, total employment, 2025 (Türkiye 2021): https://data-explorer.oecd.org/vis?df[ds]=dsDisseminateFinalDMZ&df[id]=DSD_HW%40DF_AVG_ANN_HRS_WKD&df[ag]=OECD.ELS.SAE
Who is affected
Agent-capable jobs compared with all assessed jobs that are still online on the source portals. On the home page, the entry-level share is also applied to the OECD extrapolation.
Share of agent-capable jobs by sector
| Sector of the employer | Assessed jobs online | Agent-capable | Share |
|---|---|---|---|
| internet, IT & telecoms | 1,116 | 268 | |
| electronics & automation | 767 | 151 | |
| public sector & associations | 341 | 50 | |
| banking, finance & insurance | 1,507 | 215 | |
| consulting | 244 | 32 | |
| mechanical & plant engineering | 1,562 | 177 | |
| education | 157 | 15 | |
| health & social care | 302 | 29 | |
| law, tax & business services | 1,200 | 108 | |
| advertising, marketing & PR | 292 | 26 | |
| energy & environment | 428 | 37 | |
| retail & trade | 523 | 37 |
Snapshot of 2026-10-28. Entry level: title or requirements mention junior, trainee, graduate, internship, apprenticeship, or no or first work experience. Degree: requirements mention a university or college degree. Pay: stated monthly gross minimum (else maximum) × 14, only jobs that state a salary. Sectors with at least 150 assessed jobs online. Keyword detection is approximate.
Limits
- The assessment is done by a language model, not by people. It is a strong, transparent indicator, not a verdict; borderline cases can tip either way between runs.
- New jobs arrive in assessment runs, not as a live stream. The figures change after each run.
- There is no measured monthly series of the AI quota yet, so botjobs shows no trend. It will be built from historical job ads. The API still returns a modelled series, marked estimated=true; it is not a measurement.
- The extrapolation is an estimate with the caveats listed above.
Updates
The analysed jobs per portal are updated after every assessment run, the vacancy and wage base when new official figures are published (at least quarterly). Every change to the method is recorded on this page.
- 26 Oct 2026
Counter "jobs killed by AI" (worldwide model) and the cost of one lost job (taxes, unemployment benefits, consumption, saving, second round).
- 26 Oct 2026
Taxes & social contributions, pension contributions and working hours added to the extrapolation; who is affected (entry level, degree, pay, sectors).
- 25 Oct 2026
Extrapolation to Austria and the OECD countries added; English as main language.
Data for humans and agents
- /api/stats – all figures as JSON, including the extrapolation
- /api/jobs – all listed jobs with score and reason (German and English)
- /llms.txt – short description for AI agents
Questions or press: hello@botjobs.org