What the reports rest on

The first BLS projections cycle to publish an AI-exposure classification for every detailed occupation. 831 occupations are placed in low, moderate, high or very-high exposure tiers by combining three theoretical exposure measures (Felten–Raj–Seamans; Eloundou et al.; Eisfeldt et al.) with two observed-usage measures (Anthropic Claude and Microsoft Copilot). Published alongside the 2025–35 employment projections.

• 01

  Total US employment projected to grow 3.5% (5.9 million jobs) over 2025–35.

• 02

  Office and administrative support is the fastest-declining major group (−4.0%, −752,100 jobs), with AI-powered automation cited as a driver.

• 03

  206 occupations sit in the 'very high' exposure tier; 139 of them are still projected to grow, so exposure is not a forecast of job loss.

• 04

  Fastest-declining detailed occupations include word processors and typists (−34%), data entry keyers (−26%) and telemarketers (−21%). Customer service representatives are projected to fall 5.3%.

• 05

  Software developers (+10.2%), data scientists (+34.6%) and nurse practitioners (+41%) are all highly exposed and all growing.

Uses ADP payroll data on millions of US workers through June 2026 to test whether AI is showing up in employment. The headline result is a widening gap for early-career workers in exposed occupations, with no economy-wide displacement.

• 01

  No evidence of widespread, economy-wide job displacement from AI.

• 02

  Employment of 22–25-year-olds in highly exposed occupations is 19% below where it would be had it kept pace with less-exposed peers, up from 15% in the July 2025 data.

• 03

  The effect works through reduced hiring, not firing, and is concentrated where AI use substitutes for tasks; where use is complementary, employment is flat or rising.

• 04

  Experienced workers show no comparable gap. Adjustment is happening through employment rather than pay.

Sixth Economic Index release. Links a survey of roughly 9,700 Claude users to privacy-preserving usage data, and publishes an 'observed exposure' figure per occupation: the share of an occupation's O*NET tasks already being done with Claude.

• 01

  AI use is concentrated in computer, mathematical and management occupations; physical occupations remain under-represented.

• 02

  Close to six in ten respondents expect AI to be able to do more of their work next year than today; over a third expect it to handle most or nearly all of their tasks.

• 03

  Automation (directive or feedback-loop use) rises with delegation and agentic tools such as Claude Code; augmentation still dominates on Claude.ai.

• 04

  Observed exposure is highest for computer programmers (0.75), database architects, software QA testers and customer service representatives.

The UK government's first consolidated assessment of how AI capability trends are translating into labour-market effects, and of the gaps in the evidence.

• 01

  Around 70% of UK workers are in occupations containing tasks AI could perform or enhance, above the US and advanced-economy average of about 60%; roughly half of those are in roles where AI is more likely to complement than replace.

• 02

  A one-standard-deviation increase in AI exposure was associated with a 3.9% fall in UK job postings, becoming significant about seven months after ChatGPT's release; McKinsey found UK adverts fell 38% for high-exposure occupations against 21% for low-exposure ones (2022–25).

• 03

  56% of firms using AI report productivity gains, mostly self-assessed at up to 20%; causality between exposure and hiring declines is not established.

Analyses 200,000 anonymised Bing Copilot conversations, classifies the work activities people ask AI to do and how successfully it does them, then maps those activities to O*NET occupations to produce an 'AI applicability score' for 785 occupations.

• 01

  Highest applicability in knowledge work: interpreters and translators (0.49), historians, writers, customer service representatives, sales representatives of services.

• 02

  Lowest applicability in physical work: dredge operators, roofers, maids, massage therapists and most construction trades.

• 03

  The most common activities people seek AI help with are gathering information and writing; the most common activities AI performs are providing information, writing, teaching and advising.

• 04

  The authors stress that applicability measures where AI is useful for sub-tasks, not whether a job will be replaced.

Updates the ILO's 2023 global index using 52,558 task-level judgements from workers and experts, then scores every ISCO-08 occupation into four exposure gradients.

• 01

  One in four workers worldwide is in an occupation with some generative-AI exposure; 3.3% of global employment is in the highest gradient, rising to 34% of all employment having some exposure in high-income countries.

• 02

  Clerical occupations remain the most exposed; all 13 gradient-4 occupations are clerical, including data entry clerks, typists, accounting and bookkeeping clerks and general office clerks.

• 03

  The 2025 scores are lower than 2023 at the top: the highest task score is 0.76 and the highest occupational mean 0.70, reflecting two years of real experience with the tools.

• 04

  Some strongly digitised professional and technical occupations have increased exposure; transformation, not elimination, is the most likely effect.

Survey of over 1,000 employers representing 14 million workers across 22 industries and 55 economies, on expected job creation, displacement and skill change to 2030.

• 01

  Employers expect 170 million jobs to be created and 92 million displaced by 2030, a net gain of 78 million (7% of employment).

• 02

  86% of employers expect AI and information-processing technology to transform their business by 2030; 39% of workers' skills are expected to change.

• 03

  Fastest-growing roles: big data specialists, fintech engineers, AI and machine-learning specialists, software and applications developers.

• 04

  Fastest-declining roles: cashiers and ticket clerks, administrative assistants and executive secretaries, printing workers, accountants and auditors, data entry clerks, bank tellers. Graphic designers and legal secretaries appear on the declining list for the first time.

Annual assessment of OECD labour markets, with a chapter on young entrants and a review of whether large language models explain their difficulties.

• 01

  OECD unemployment was 4.9% in May 2026 with employment at a record 72.1%, but signs of weakening are emerging.

• 02

  Young graduates' unemployment gap has been widening since before the pandemic in every country analysed; the role of LLMs so far appears limited.

• 03

  US productivity has been lifted by sectors central to AI adoption; AI adoption amplifies the importance of non-cognitive skills and is reducing some returns to formal education.

Reviews evidence on entry-level hiring across the US, UK and Sweden and sets out a framework for safeguarding early-career pathways as AI absorbs junior tasks.

• 01

  Entry-level hiring slowdowns are evident and most pronounced in AI-exposed occupations, but declines began before ChatGPT and have several causes.

• 02

  Cites a 16% decline in entry-level jobs in AI-exposed fields in the US since late 2022.

• 03

  Without deliberate support, AI risks reinforcing existing inequalities in who gets a first job.

Analysis of over one billion job adverts in 27 countries plus firm financials, tracking how AI exposure relates to productivity, wages and hiring.

• 01

  Wage premium for AI skills rose to 62% (from 57%); jobs requiring AI skills grew 69% against 9% for the market overall.

• 02

  Firms in the most AI-exposed sectors recorded 34% productivity growth since 2018 versus 24% for the least exposed; the top fifth of exposed firms reached 163%.

• 03

  A two-track market is emerging: jobs 'professionalised' by AI grow twice as fast, with 42% faster wage growth, than jobs 'democratised' by it.

• 04

  The most exposed roles are adding tasks that rely on empathy, judgement and creativity 2.5 times faster than the least exposed.

Annual compendium of adoption, investment and labour-market indicators, with labour data from Lightcast and the Stanford Digital Economy Lab.

• 01

  Organisational AI adoption reached 88% of surveyed firms in 2025; generative AI is used in at least one business function at 70%.

• 02

  Generative AI reached 53% population-level adoption within three years, faster than the PC or the internet.

• 03

  AI skills now appear in 2.5% of US job postings, up 55% in a year; mentions of agentic-AI skills rose over 280%.

• 04

  Employment of software developers aged 22–25 has fallen nearly 20% since 2024; one third of organisations expect AI to reduce headcount in the coming year.

Fifth release. Tracks how usage patterns changed between November 2025 and February 2026, including the migration of coding work from augmentative chat use to automated API workflows.

• 01

  About 49% of occupations have now seen at least a quarter of their tasks performed with Claude, up from 36% in January 2025.

• 02

  Coding tasks continue to migrate from Claude.ai to automated first-party API workflows, where directive use dominates.

• 03

  Customer service tasks (payments, billing support) are prevalent in API traffic, pointing to higher real-world exposure for customer service representatives.

Examines demand and supply of new IT and AI skills across advanced and emerging economies using job-vacancy data.

• 01

  About one in ten vacancies in advanced economies now demands at least one new skill; vacancies demanding AI skills post higher wages.

• 02

  Diffusion of AI skills is linked to lower employment in occupations with high exposure and low complementarity to AI, posing particular challenges for young people.

• 03

  Builds on the IMF's 2024 finding that about 40% of global employment (60% in advanced economies, roughly 70% in the UK) is exposed to AI.

Refreshes MGI's 2017 automation model with 2025 technology performance across 18 human capabilities, 800 occupations and 2,000 work activities.

• 01

  Currently demonstrated AI agents and robots could in theory automate about 57% of US work hours (agents ~44%, robots ~13%); this is technical potential, not a forecast.

• 02

  Roughly 40% of the workforce is in occupations where automatable activities make up more than half of working hours, concentrated in legal, administrative and other 'agent-centric' work.

• 03

  A midpoint adoption scenario puts about 27% of US work hours automated by 2030, with $2.9 trillion of value at stake.

• 04

  Demand for AI fluency grew sevenfold in two years, faster than any other skill in US postings; skills tied to assisting and caring will change least.

Introduces the GenAI Skill Transformation Index, scoring almost 2,900 skills from 53.5 million US postings by how far generative AI could transform them.

• 01

  46% of skills in a typical posting are open to hybrid or full transformation; 26% of jobs are poised to transform radically.

• 02

  Software development is the most exposed occupation (81% of typical skills rated hybrid); nursing the least (68% minimal).

• 03

  Follow-up work (July 2026) finds software postings up almost 15% since early 2025, concentrated in senior and AI-titled roles, and advertised pay rising fastest in exposed occupations (September 2026).

Regularly updated analysis of US household survey microdata for any AI footprint in employment, unemployment and occupational mix.

• 01

  The occupational mix is changing only about one percentage point faster than during the internet era, and the shift predates ChatGPT.

• 02

  Measures of AI exposure, automation and augmentation show no clear relationship to changes in employment or unemployment so far.

• 03

  A synthetic difference-in-differences design finds no statistically distinguishable effect on exposed occupations' employment or real wages yet.

Uses UK administrative data to decompose the rise in young people not in education, employment or training.

• 01

  The NEET rise has been similar for graduates and non-graduates, cutting against the idea that AI is disproportionately hitting entry-level graduate jobs.

• 02

  Youth minimum-wage rises, employer National Insurance changes, AI adoption and declining youth mental health are all plausible structural contributors.

Survey of 20,000 AI-using workers in 10 countries plus trillions of Microsoft 365 productivity signals, focused on how far organisations have got with agents.

• 01

  Active agents in the Microsoft 365 ecosystem grew 15x year on year (18x in large enterprises).

• 02

  Only 19% of AI users are 'frontier' users with both strong individual skill and organisational support; 16% are stalled.

• 03

  Organisational factors (culture, manager support, talent practices) account for twice the reported AI impact of individual effort.

Two 2026 research notes estimating AI's current and prospective effect on US employment, separating substitution from augmentation.

• 01

  Base case: 6–7% of US workers displaced over a roughly ten-year adoption period (range 3–14%), with a peak unemployment increase of about 0.6 points.

• 02

  AI has already reduced monthly payroll growth by about 16,000 jobs over the past year: roughly 25,000 lost to substitution, 9,000 added through augmentation.

• 03

  Negative effects fall mainly on younger, less experienced workers in substitutable occupations; a one-standard-deviation rise in substitution exposure widens the entry-to-experienced wage gap by about 3.3 points.