New jobs, business growth and workforce transitions.
Research reviewed 13 September 2026New AI-related opportunities globally in 2023–2025, reported by LinkedIn.
Platform analysis, not a census or a net balance after job losses. Several of these occupations existed before the current AI boom.
What does this mean for existing workers? →Examples named in LinkedIn’s report; opportunities span model development, data work and implementation.
Change over one year · LinkedIn’s January 2026 report
100+ manually classified firms, including startups and Big Tech. A selected company sample, not growth across the entire labour market.
2025 · OECD analysis of Preqin deals
$258.7B of $427.1B; rounded share. Funding is not revenue, profit or job creation.
Worldwide in 2025, according to Stanford’s AI Index.
Selected countries. Funded companies are not newly incorporated firms or individual founders. Quid data; definitions differ from OECD/Preqin.
Selected years · USD billions
OECD / Preqin. Selected published years, without interpolation.
Funding pays for the capacity to build businesses; employment and productivity tell us more about what follows. See AI-company hiring → See business outcomes →
Issues per hour · comparison without AI = 100
15% higher productivity across 5,172 agents at one firm.
Field study published in 2025; earlier data. Estimated effect in this setting, not a universal gain.
Completed tasks · without AI = 100
26% more completed tasks across three randomized field experiments with 4,867 developers.
Microsoft, Accenture and a Fortune 100 firm; 2025 paper. Pooled estimate 26.08%, standard error 10.3 percentage points. Each trial was noisy.
These studies measured output on specific tasks. They do not show that every worker, model or workflow becomes more productive.
Time per task · comparison without AI = 100
19% more time in a trial with 16 developers and 246 tasks.
Early-2025 tools; reported interval +2% to +39%. This is a historical result, not a current coding-agent benchmark.
The smaller METR trial used experienced open-source developers working in familiar repositories. The larger experiments tested coding-assistant access in companies. Different tasks, tools and participants can produce different results.
METR’s February 2026 follow-up suggests improvement, but cannot establish a reliable current estimate.
Share of GenAI-using SMEs reporting each benefit
OECD survey of 5,000+ SMEs in seven countries. The 26% is the share reporting higher revenue, not a 26% revenue increase. Self-reports do not establish causation.
Projected increase in U.S. GDP level from GenAI
Faster work can increase output from the same resources. Penn Wharton models how that could accumulate.
Future GDP levels relative to a baseline without the modelled GenAI contribution. These are not annual growth rates or gains already realized; adoption and cost-saving assumptions matter.
A randomized trial of an AI business adviser in Kenya did not establish an average revenue or profit benefit. Effects differed by entrepreneurs’ starting performance. Access to a tool is only one part of making it useful.
of employment has some GenAI exposure.
Global employment
1 person = 1% of employment. ILO’s 2025 model measures potential task exposure, not jobs lost; 3.3% globally is in the highest category, within the 25%.
Employment for 22–25-year-olds in AI-exposed U.S. occupations fell behind less-exposed peers, mainly through weaker hiring.
Stanford ADP study, revised August 2026. Observational, not proof AI caused the gap; education controls weaken some patterns.
WEF forecast, 2025–2030. Includes all major economic trends, not AI alone. These are not jobs already created.
These studies do not establish a global net-jobs figure caused by AI. ILO ↗ WEF ↗
Ask which tasks they want help with and where human judgment should remain. Stanford’s WorkBank research documents preferences; OECD surveys associate consultation and training with better reported outcomes.
Provide training time, routes into new roles and support for affected workers. Track junior hiring alongside productivity. These are recommendations, not guaranteed fixes.
Measure time, accuracy and rework before expanding. Harvard’s consulting research found AI helped on some tasks and reduced correctness on a task beyond its capabilities.
Paraphrased views, not quotations or a claim of consensus.
Test AI on the organization’s own tasks before assuming it will help.
Primary university studies, government and intergovernmental reports, ethical-AI research and company disclosures published from September 2021–September 2026. This is a curated guide, not an exhaustive systematic review. Older underlying data are labelled.
Estimates, forecasts, surveys and observed results are kept distinct. Company claims are identified as disclosures. Charts are redrawn from cited data, with rounded values and calculated shares labelled; no missing years are interpolated.
Industry comparisons show scope and date because electricity, onsite water and supply-chain footprints are different measures. Findings do not automatically generalize beyond the studied people, products or settings.
Each “Notes” button gives the finding, scope and limits.
Download source ledger ↗International Energy Agency
International Energy Agency
Lawrence Berkeley National Laboratory
Google researchers
Luccioni, Viguier & Ligozat
Wang et al. · Nature Computational Science
International Labour Organization
World Economic Forum
Stanford Digital Economy Lab
Brynjolfsson, Li & Raymond · Quarterly Journal of Economics
Stanford HAI
Otis et al. · Management Science / Berkeley Haas
International Telecommunication Union
Lund University · The Lancet
Hofmann et al. · Nature
Costello, Pennycook & Rand · Science / MIT Sloan
National Institute of Standards and Technology
Google · Michael Terrell
Microsoft · Brad Smith & Melanie Nakagawa
Microsoft · Steve Solomon
Microsoft · Scott Guthrie
Amazon · Tyler Greenawalt
McDonald et al. · MIT Lincoln Laboratory and collaborators
MIT Han Lab · Yang et al.
Luccioni, Jernite & Strubell · FAccT
Xu et al. · Princeton / Joule
Science Based Targets initiative
Science Based Targets initiative
Harvard Business School · Rachel Layne / McFowland et al.
Ada Lovelace Institute & Alan Turing Institute
University of Melbourne & KPMG · Gillespie, Lockey and colleagues
Andrew A. Chien · Stanford Energy Seminar
MIT Energy Initiative · Emma Strubell and symposium participants
AI Now Institute
International Energy Agency
U.S. Energy Information Administration
U.S. Geological Survey
European Environment Agency
United Nations Statistics Division
International Energy Agency
Sony Interactive Entertainment
Ofcom · commissioned research by Carnstone
LinkedIn Economic Graph
Cui, Demirer, Jaffe, Musolff, Peng & Salz · Microsoft Research
Penn Wharton Budget Model · University of Pennsylvania
Nearing et al. · Google Research and collaborators · Nature
EMBL European Bioinformatics Institute
International Energy Agency
No matching sources. Try another word or topic.