Goldman Sachs predicts a 7% rise in global GDP, about $7 trillion, and 1.5% a year more US productivity growth over ten years: the headline projection the rest of this page is measured against.
Goldman Sachs 2023, cited in Acemoglu (MIT Economics), 2025
CorroboratedAround 80% of the US workforce could have at least a tenth of their tasks affected by large language models, and about 19% at least half.
Eloundou et al. (OpenAI), 2023
Almost 40% of global employment is exposed to AI, replacing some work and complementing the rest, with advanced economies facing both more risk and more opportunity.
IMF (Georgieva), 2025
Only 5% of US businesses use AI in producing goods and services, though about 28% of individuals say they use generative AI for their job.
Congressional Budget Office, 2025
Annual value, measured productivity, record investment and adoption: the gains already counted.
McKinsey estimates generative AI could add $2.6–4.4 trillion a year, with further productivity growth depending on how fast it is adopted and labour redeployed.
McKinsey, cited in Bayraktar, 2026
CorroboratedAcross customer-support agents, access to a GPT-based assistant raised productivity by 14% on average, with much larger gains for novice and less-skilled agents.
Brynjolfsson et al., cited in Brookings Institution, 2025
CorroboratedCorporate AI investment reached $252.3 billion in 2024, with private investment up 44.5%; total AI investment has grown more than thirteenfold since 2014.
Stanford HAI, AI Index 2025, 2025
Corroborated78% of organisations reported using AI in 2024, up from 55% the year before, while research increasingly finds AI raises productivity and narrows skill gaps.
Stanford HAI, AI Index 2025, 2025
CorroboratedKlarna reported its AI system doing the work of 700 customer-service agents about a year after it laid off 700 employees.
Brookings Institution, 2025
Company-reportedThe share of US jobs in science, technology, engineering and maths grew from 6.5% in 2010 to nearly 10% in 2024: an almost 50% increase as firms hire technical talent.
Harvard Gazette, on Deming et al., 2025
Single-sourceAutomation has raised wages before: falling labour share accounted for 16% of US real wage growth from 1954 to 2019, and the authors find further automation should raise wages in all twelve countries studied.
Autor et al., 2026
Peer-reviewedSpending on AI software, hardware and services is projected to reach $521 billion by 2027, against $180 billion in 2023.
Bank of America, 2025
Company-reportedNo verified figure exists for how many of the jobs AI creates go to the people whose jobs it displaces. Forecasts count both totals; none tracks the same workers across the switch.
No verified figure exists for productivity gains outside the handful of firms that published field studies. Most adopters report usage, not output.
Who is exposed, how big the gain really is, and what happens when everyone writes the same polished application.
In UK data, roughly 60% of workers at the 90th income percentile are in occupations where AI can do a large share of tasks, against 15% at the 10th percentile. High earners also hold the most capital.
IMF working paper, 2025
Single-sourceBuilding the forecast up from tasks, Acemoglu finds total-factor productivity about 0.71 percentage points higher in ten years and GDP about 1.1% higher: far below the 7% of the headline forecast.
Acemoglu (MIT Economics), 2025
Single-sourceExposure to automation by generative AI is concentrated in richer countries; augmentation reaches more jobs everywhere, but infrastructure gaps may widen the productivity divide.
ILO (Gmyrek, Berg & Bescond), 2023
Single-sourceWhen an AI writing tool let every freelancer write a polished pitch, employers could no longer tell strong from weak applicants: average wages fell 5% and hiring per job post 1.5%.
Galdin & Silbert, 2025
Single-source46% of young Americans think it at least somewhat likely that AI will replace their job in the next five years, and roughly two-thirds want government to act against AI job losses.
Brookings Institution, 2025
Single-sourceNo verified figure exists for how much of the forecast GDP gain reaches wages rather than profits. The models that estimate it disagree on the share by design.
No verified count exists of layoffs caused by AI. Employers' stated reasons are not audited, and AI is often one factor among several.
Jobs exposed, inequality widened, incomes squeezed and routine work thinned out.
Up to 300 million full-time jobs could be automated globally, with a quarter of exposed positions at risk of replacement by generative AI.
Hatzius (Goldman Sachs) 2023, cited in Ahmadi, 2024
CorroboratedOver the last 40 years, about 50–70% of the increase in US wage inequality has been attributed to new automation technologies.
Brookings Institution, 2025
CorroboratedA system-dynamics model finds a moderate rise in AI capital relative to labour could double underutilisation and cut disposable income per person by 26% by mid-2050, unless job creation rises more than tenfold.
Scientific Reports, 2025
Single-sourceBetween 2013 and 2023 retail sales jobs fell from 7.5% to 5.7% of the US job market, a 25% drop, as early adopters of predictive AI reshaped e-commerce.
Harvard Gazette, on Deming et al., 2025
Single-sourceClerical work is the one broad group highly exposed: 24% of its tasks are highly exposed and a further 58% at medium level, 82% in all.
ILO (Gmyrek, Berg & Bescond), 2023
Single-sourceUS private AI investment reached $109.1 billion in 2024, nearly 12 times China's $9.3 billion and 24 times the UK's $4.5 billion.
Stanford HAI, AI Index 2025, 2025
CorroboratedChina and North America will account for about 70% of AI's global economic impact by 2030; PwC puts China's GDP boost at about 26% and North America's at 14.5%.
Bank of America, 2025
Company-reportedWorkers' share of US income fell from 67.8% to 58.4% between 1987 and 2019, with technological change a leading explanation.
Autor et al., 2026
Peer-reviewedA 15% capital tax could cut the wage Gini by 3.4 points and the wealth Gini by 3.7, but would reduce output by 26.9% and average wages by 11.8%.
IMF working paper, 2025
Single-sourceA sevenfold spread in forecasts, one exposed occupation in nine, and counts that measure different things.
Forecasts of what AI adds to output over a decade run from about 1.1% (Acemoglu, built up from tasks) to 7% (Goldman Sachs). The spread is the finding.
Acemoglu (MIT Economics), 2025
CorroboratedOnly clerical work is highly exposed; in every other broad group, at most 1–4% of tasks are highly exposed. The main effect is likely augmentation, not automation.
ILO (Gmyrek, Berg & Bescond), 2023
Single-sourceOnly 5% of US businesses incorporate AI into production in more than an incidental way, yet about 28% of individuals say they use generative AI for their job: the two counts measure different things.
Congressional Budget Office, 2025
CorroboratedThe most cited exposure scores were set against models of early 2023; frontier capability has since moved about 26% beyond them, so the standard map of exposed work is already stale.
arXiv study of creative professions, 2025
Single-sourceA 2013 study put 47% of US occupations at imminent risk from computers; the measured pace of occupational change slowed instead, until it picked up again from 2019.
Harvard Gazette, on Deming et al., 2025
Single-sourceProjections are not outcomes. Where sources disagree, both are printed here; where a figure is company-reported, contested or single-source, it says so. Every statistic carries its origin and its confidence label.
Goldman Sachs and McKinsey forecasts are cited via Acemoglu and Bayraktar, and the customer-support study via Brookings, as those sources state them.
