Will AI actually boost productivity and growth?
The numbers
What the published studies estimate on “Will AI actually boost productivity and growth?”
77 numbers from 52 published papers, each taken from the source’s own words (its abstract, or a verbatim passage of a curated source) with the sentence it comes from. Hover or tap a point for the quote; every point links to its paper.
Forecasts of AI’s effect on growth
Each point is one estimate of extra growth a year from AI, by the date it was published; bars show the author’s own range. 34 estimates from 32 sources, grouped as in Tom Cunningham’s Forecasts of AI & Economic Growth: economists and AI insiders. Measures (GDP, TFP, labour productivity), regions and horizons differ, and a few are converted from a level or a survey (the tooltip says how), so read the spread, not a single number. Log scale.
EconomistsAI insidersShaded: the middle half of each groupProductivity gain (including time saved)Open marker: AI labs’ own research
Who gains more from AI at work
Studies that measured AI’s effect on output separately for less and more experienced (or lower- and higher-performing) workers. When the less experienced gain more, the gap between workers narrows; when the better performers gain more, it widens. 1 study, each number with its quote; settings differ.
Ghorbani+19% more experienced or higher performers
0%20%
Less experienced or lower performersMore experienced or higher performersGainLoss
Effects measured on tasks and jobs
Change in time, output or quality in a study setting. A negative time change means faster work.
Claims settlement time at ZhongAn after AI automationWANG & SOLOV'EVA · −87.5%
Completion time on matched knowledge-work tasks with an AI agentYang et al. · −87%
Time to implement an HTTP server with GitHub CopilotPeng et al. · −55.8%
Coding lead time with AI assistanceJethwa · −55.8%
Time on professional writing tasks with ChatGPTNoy & Zhang · −40%
Task time for returning developers using AI (later study; selection effects)METR · −18% (range −38 to 9)
Primary care, AI scribe: note-writing time after 150 daysWaken et al. · −16% (range −18 to −13)
Task time for newly recruited developers using AI (later study; selection effects)METR · −4% (range −15 to 9)
Primary care, AI scribe: work relative value units after 150 daysWaken et al. · +2%
Duolingo, GitHub Copilot: engineering velocity in a familiar environmentOhinok & Лесюк · +10%
Customer-support issues resolved per hour with an AI assistantBrynjolfsson et al. · +15%
Output quality on professional writing tasks with ChatGPTNoy & Zhang · +18%
Task time for experienced open-source developers using AI (early 2025 RCT)METR · +19% (range 2 to 39)
Duolingo, GitHub Copilot: engineering velocity in new or unfamiliar repositoriesOhinok & Лесюк · +25%
Tasks completed by developers with GitHub CopilotCui et al. · +26%
Gain in task performance with AI on tasks beyond IT apprentices' training (percentage points)Brosch et al. · +28.5% (range 26 to 31)
Law-student trial: productivity gains with a legal AI tool and a reasoning modelGhorbani · +90% (range 50 to 130)
−100%150%
Productivity gain (including time saved)No gain, or a lossOpen marker: AI labs’ own research
Effects on the level of output or productivity
Cumulative change in a level over the stated horizon, not a yearly rate.
TFP gain over 10 years (upper bound)Acemoğlu · +0.66%
Cumulative productivity gain over five yearsMisch et al. · +1%
Nonfarm business output level gain by 2036Office · +1%
Productivity and GDP level gain by 2035Arnon · +1.5%
TFP level gain over about ten yearsBergeaud · +2.9%
Productivity and GDP level gain by 2075Arnon · +3.7%
GDP per person level gain by 2050 (model simulation)Jones & Tonetti · +4%
Global GDP level gain over 10 yearsBriggs & Kodnani · +7%
Productivity and output level gain over 10 years (Level scenario)Baily et al. · +18%
GDP level gain after 10 yearsAldasoro et al. · +30%
0%30%
Productivity gain (including time saved)
Effects reported as multiples
How many times faster or more output, as reported.
Time savings factor on Claude Code-assisted tasks, 7 METR staff (author calls it a soft upper bound on uplift)METR · 1.5× (range 1.5 to 13)
Self-reported change in the value of technical workers' work due to AI (median)METR · 1.7× (range 1.4 to 2)
Modelled researcher output uplift from coding agents at Anthropic (Cobb-Douglas; one author's view)METR · 2.83×
Employee productivity at insurer ZhongAn after AI automation (2019-2024)WANG & SOLOV'EVA · 30×
0×30×
Productivity gain (including time saved)No gain, or a lossOpen marker: AI labs’ own research
Shares
Shares of jobs, tasks, workers or firms, as reported. Definitions differ between studies.
Tasks that could be done significantly faster with an LLM aloneEloundou et al. · 15%
Share of the AI-growth effect that runs through industrial development, 36 countries, 2010-2022Gyau et al. · 23.6%
Tasks that could be done significantly faster with LLM-powered softwareEloundou et al. · 51.5% (range 47 to 56)
0%60%
Other effects
Changes in employment, wages, adoption and other measures, as reported.
Code churn with ungoverned AI code generationJethwa · +19.4%
0%20%
Productivity gain (including time saved)
Read the source numbers
Each row retains its own definition, population and horizon. Rows are not a pooled estimate; a shared percentage unit does not make them comparable.
| Source | What it measures | Reported value | Source quotation |
|---|---|---|---|
| The simple macroeconomics of AI | TFP gain over 10 years (upper bound) US · 10 years | +0.66% | no more than a 0.66% increase in total factor productivity (TFP) over 10 years |
| The simple macroeconomics of AI | TFP gain over 10 years, adjusted for hard-to-learn tasks (upper bound) US · 10 years | +0.53% | predicted TFP gains over the next 10 years are even more modest and are predicted to be less than 0.53% |
| The simple macroeconomics of AI | Extra annual TFP growth US · 10 years | +0.066 pp a year (range 0.053 to 0.066) | no more than a 0.66% increase in total factor productivity (TFP) over 10 years |
| AI and Growth: Where Do We Stand? | Extra annual aggregate productivity growth (past-revolutions approach) 10 years | +1.05 pp a year (range 0.8 to 1.3) | the AI revolution should increase aggregate productivity growth by between 0.8 and 1.3pp per year over the next decade |
| AI and Growth: Where Do We Stand? | Extra annual TFP growth, median (task-based approach) 10 years | +0.68 pp a year | we obtain a median estimate of 0.68pp additional annual total factor productivity (TFP) growth |
| Sizing the prize: What's the real value of AI for your business and how can you capitalise? | Extra annual global GDP growth, implied by the 2030 level gain World · 2017-2030 | +1 pp a year | Global GDP will be up to 14% higher in 2030 as a result of the accelerating development and take-up of AI |
| Sizing the prize: What's the real value of AI for your business and how can you capitalise? | Global GDP level gain by 2030 (upper bound) World · by 2030 | +14% | Global GDP will be up to 14% higher in 2030 |
| Situational Awareness: The Decade Ahead | Extra annual GDP growth from AI Region and horizon not recorded | +30 pp a year | We could see economic growth rates of 30%/year and beyond, quite possibly multiple doublings a year. |
| World Economic Outlook Update, January 2026: Global Economy: Steady amid Divergent Forces | Extra annual global growth in the medium term World · medium term | +0.45 pp a year (range 0.1 to 0.8) | between 0.1 and 0.8 percentage points per year in the medium term |
| The $10 trillion question: Financing the AI buildout | Extra annual GDP growth from AI U.S. | +3.63 pp a year | A new BPEA paper projects that this buildout will total $10.3 trillion from 2025 to 2032, an average of 3.63% of U.S. |
| Artificial Intelligence and Economic Growth in Jordan: A Mixed-Methods Study of Commercial Banking Applications as a Mediating Channel | Extra annual GDP growth from AI Hashemite Kingdom of Jordan | +0.83 pp a year | The total effect is 0.829 at a significance level of 0.000, and the model accounts for 65.9% of the variance in economic growth. |
| The Potentially Large Effects of Artificial Intelligence on Economic Growth | Extra annual labour productivity growth US · 10 years | +1.5 pp a year | ~+1.5pp/yr US labor productivity |
| The Potentially Large Effects of Artificial Intelligence on Economic Growth | Global GDP level gain over 10 years World · 10 years | +7% | ~+7% global GDP over 10y |
| The Economic Potential of Generative AI: The Next Productivity Frontier | Extra annual productivity growth potential World | +0.35 pp a year (range 0.1 to 0.6) | ~+0.1-0.6pp/yr productivity growth potential |
| Scenarios for the Transition to AGI | Steady-state output growth in the baseline AGI scenario (a total rate) Region and horizon not recorded | +18 pp a year | the economy remains in region 2 where labor and capital are perfectly substitutable, with steady-state growth of 18% per year |
| Miracle or Myth? Assessing the macroeconomic productivity gains from Artificial Intelligence | Extra annual TFP growth US · 10 years | +0.42 pp a year (range 0.25 to 0.6) | annual aggregate total-factor productivity growth due to AI range between 0.25-0.6 percentage points |
| Miracle or Myth? Assessing the macroeconomic productivity gains from Artificial Intelligence | Extra annual labour productivity growth US · 10 years | +0.65 pp a year (range 0.4 to 0.9) | (0.4-0.9 pp. for labour productivity) |
| Macroeconomic productivity gains from Artificial Intelligence in G7 economies | Extra annual labour productivity growth G7, high AI exposure (e.g. US, UK) · 10 years | +0.85 pp a year (range 0.4 to 1.3) | annual aggregate labour productivity growth due to AI range between 0.4-1.3 percentage points in countries with high AI exposure |
| Machines of Mind: The Case for an AI-Powered Productivity Boom | Productivity and output level gain over 10 years (Level scenario) US · 10 years | +18% | generative AI raises the level of productivity and output by an additional 18% over ten years |
| Machines of Mind: The Case for an AI-Powered Productivity Boom | Extra annual growth in the Level+Growth scenario: the 18% level gain as a yearly rate plus 1 pp US · 10 years | +2.7 pp a year | additionally includes a one percentage point boost in the rate of growth over the baseline rate |
| The Past, Present and Future of European Productivity | TFP level gain over about ten years euro area · 10 years | +2.9% | We predict TFP gains of 2.9% in the medium run (say in the next ten years) in the euro area |
| The Past, Present and Future of European Productivity | Extra annual TFP growth euro area · 10 years | +0.29 pp a year | equivalent to an additional 0.29 percentage points per year |
| AI and Productivity in Europe | Cumulative productivity gain over five years Europe · 5 years | +1% | the medium-term productivity gains for Europe as a whole are likely to be modest, at around 1 percent cumulatively over five years |
| AI and Productivity in Europe | Extra annual productivity growth Europe · 5 years | +0.2 pp a year | at around 1 percent cumulatively over five years |
| Jack Clark on AI's Uneven Impact (Ep. 242) | Extra annual growth, bear to bull case Region and horizon not recorded | +4 pp a year (range 3 to 5) | my bear case on all of this is 3 percent, and my bull case is something like 5 percent |
| The Projected Impact of Generative AI on Future Productivity Growth | Productivity and GDP level gain by 2035 US · 2025-2035 | +1.5% | AI will increase productivity and GDP by 1.5% by 2035 |
| The Projected Impact of Generative AI on Future Productivity Growth | Productivity and GDP level gain by 2075 US · 2025-2075 | +3.7% | nearly 3% by 2055, and 3.7% by 2075 |
| The Projected Impact of Generative AI on Future Productivity Growth | Extra annual growth to 2035 US · 2025-2035 | +0.15 pp a year | AI will increase productivity and GDP by 1.5% by 2035 |
| The economic impact of AI | Extra annual US growth (best guess) US | +0.38 pp a year (range 0.25 to 0.5) | will boost the annual US growth rate by one-quarter to one-half of a percentage point |
| Why I think AI take-off is relatively slow | Extra annual growth Region and horizon not recorded | +0.5 pp a year | AI will boost economic growth rates by half a percentage point a year |
| Experimental evidence on the productivity effects of generative artificial intelligence | Time on professional writing tasks with ChatGPT US | −40% | The average time taken decreased by 40% |
| Experimental evidence on the productivity effects of generative artificial intelligence | Output quality on professional writing tasks with ChatGPT US | +18% | output quality rose by 18% |
| Generative AI at Work | Customer-support issues resolved per hour with an AI assistant Region and horizon not recorded | +15% | as measured by issues resolved per hour, by 15% on average |
| GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models | Tasks that could be done significantly faster with an LLM alone US | 15% | about 15% of all worker tasks in the US could be completed significantly faster at the same level of quality |
| GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models | Tasks that could be done significantly faster with LLM-powered software US | 51.5% (range 47 to 56) | this share increases to between 47 and 56% of all tasks |
| The Impact of AI on Developer Productivity: Evidence from GitHub Copilot | Time to implement an HTTP server with GitHub Copilot Region and horizon not recorded | −55.8% | completed the task 55.8% faster than the control group |
| The Effects of Generative AI on High Skilled Work: Evidence from Three Field Experiments with Software Developers | Tasks completed by developers with GitHub Copilot Region and horizon not recorded | +26% | +26% completed tasks |
| How AI Agents Reshape Knowledge Work: Autonomy, Efficiency, and Scope | Completion time on matched knowledge-work tasks with an AI agent Region and horizon not recorded | −87% | Computer reduces completion time from 269 to 36 minutes on matched tasks, lowering estimated time and cost by 87% and 94% |
| Because 8 ≈ e², Anthropic's researcher uplift is plausibly >2x | Modelled researcher output uplift from coding agents at Anthropic (Cobb-Douglas; one author's view) Region and horizon not recorded | 2.83× | then researcher uplift is \(M^\beta = 2.83\) |
| Measuring the Self-Reported Impact of Early-2026 AI on Technical Worker Productivity | Self-reported change in the value of technical workers' work due to AI (median) Feb-Apr 2026 | 1.7× (range 1.4 to 2) | Participants self-reported a median 1.4–2x change in the value in their work due to AI tools |
| Measuring the Self-Reported Impact of Early-2026 AI on Technical Worker Productivity | Self-reported speed change for technical workers (median) Feb-Apr 2026 | 3× | The median self-reported speed change (which we expect to be higher than value change) is 3x |
| Estimating AI productivity gains | Extra annual labour productivity growth from current models (extrapolation, not a prediction) US · 10 years | +1.8 pp a year | current-generation AI models could increase US labor productivity growth by 1.8% annually over the next decade |
| ARTIFICIAL INTELLIGENCE AND THE PRODUCTIVITY PARADOX: MODELING STRUCTURAL TRANSFORMATION AND LABOR MARKET POLARIZATION ACROSS ECONOMIC GROUPS | Labour productivity change per 1% more automation (panel estimate) Region and horizon not recorded | +0.44% | An increase in the level of automation by 1% leads to a 0.44% increase in productivity |
| We are Changing our Developer Productivity Experiment Design | Task time for experienced open-source developers using AI (early 2025 RCT) Feb-Jun 2025 | +19% (range 2 to 39) | Our early 2025 study found the use of AI causes tasks to take 19% longer, with a confidence interval between +2% and +39% |
| We are Changing our Developer Productivity Experiment Design | Task time for returning developers using AI (later study; selection effects) from Aug 2025 | −18% (range −38 to 9) | we now estimate a speedup of -18% with a confidence interval between -38% and +9% |
| We are Changing our Developer Productivity Experiment Design | Task time for newly recruited developers using AI (later study; selection effects) from Aug 2025 | −4% (range −15 to 9) | Among newly-recruited developers the estimated speedup is -4%, with a confidence interval between -15% and +9% |
| Analyzing coding agent transcripts to upper bound productivity gains from AI agents | Time savings factor on Claude Code-assisted tasks, 7 METR staff (author calls it a soft upper bound on uplift) January 2026 | 1.5× (range 1.5 to 13) | This method estimates a time savings factor of ~1.5x to ~13x on Claude Code-assisted tasks for 7 METR technical staff in January 2026 |
| Artificial intelligence: An X-factor in a new investment regime | Extra annual productivity growth over the next decade US · next decade | +0.25 pp a year (range 0.2 to 0.3) | a productivity boost of 0.2%– 0.3% is plausible over the next decade |
| Machines of Loving Grace: How AI Could Transform the World for the Better | Annual GDP growth in the developing world, a 'dream scenario' (a total rate) Developing world | +20 pp a year | would be 20% annual GDP growth rate in the developing world |
| How much economic growth from AI should we expect, how soon? | Extra annual economic growth from AI near future | +6 pp a year (range 3 to 9) | We expect AI will provide a 3%-9% increase to economic growth per year in the near future |
| Advances in AI will boost productivity, living standards over time | Extra annual productivity growth (their 'more reasonable scenario') US · next decade | +0.3 pp a year | AI boosts annual productivity growth by 0.3 percentage points for the next decade |
| Opportunities and risks of artificial intelligence for productivity | Extra annual TFP growth in the US US · next decade | +0.5 pp a year (range 0.3 to 0.7) | AI could raise annual total factor productivity (TFP) growth by around 0.3-0.7 percentage points in the United States over the next decade |
| Past Automation and Future A.I.: How Weak Links Tame the Growth Explosion | Extra annual growth in GDP per person, implied (a model simulation, not the authors' forecast) to 2050 | +0.16 pp a year | incomes are only 4% higher in 2050 than they would be if growth remained constant rather than accelerating |
| Past Automation and Future A.I.: How Weak Links Tame the Growth Explosion | GDP per person level gain by 2050 (model simulation) by 2050 | +4% | incomes are only 4% higher in 2050 |
| The Budget and Economic Outlook: 2026 to 2036 | Extra annual TFP growth from generative AI US · 2026-2036 | +0.1 pp a year | annual TFP growth from 2026 to 2036 is 0.1 percentage point higher than it would have been without the additional contribution from generative AI diffusion |
| The Budget and Economic Outlook: 2026 to 2036 | Nonfarm business output level gain by 2036 US · by 2036 | +1% | increases the level of output in the nonfarm business sector by 1 percent in 2036 |
| Anthropic's Chief on A.I.: 'We Don't Know if the Models Are Conscious' | Developed-world GDP growth he can see AI bringing (a total rate) Developed world | +10 pp a year (range 5 to 15) | I can see a world where A.I. brings the developed world G.D.P. growth to something like 10, 15 percent. Five, 10, 15 |
| Forecasting the Economic Effects of AI | Extra annual GDP growth implied by the survey's median forecast over the baseline US · to 2030 | +0.5 pp a year | an annual GDP growth rate of 2.5%, which exceeds both the typical medium-run (2.0%) and long-run (1.7%) baseline forecasts |
| Longitudinal Expert AI Panel, Wave 6: Economic Effects of AI | Extra annual GDP growth implied by experts' forecasts over the slow-progress scenario (to rapid progress) US · 2025-2030 | +0.4 pp a year (range 0.4 to 1.7) | with median forecasts of 2.7%, 2.5%, and 2.5% annualized real GDP growth for 2025–2030, respectively—broadly in line with recent historical rates. Scenario assumptions make a substantial difference, however. Under Rapid Progress, the median expert expects 4.0% annualized growth to 2030 and 5.0% to 2050, while under Slow Progress those figures fall to 2.3% and 2.5%. |
| Global trade linkages and the cross-country distribution of the gains from AI | Extra annual growth in per capita real income, across countries Cross-country · next decade | +0.53 pp a year (range 0.1 to 0.95) | by 0.1–0.95 percentage points annually over the next decade in our central scenario |
| The impact of artificial intelligence on output and inflation | Extra annual GDP growth, implied by the 10-year level gain first 10 years | +2.7 pp a year | at which point GDP is almost 30% higher than it would have been without |
| The impact of artificial intelligence on output and inflation | GDP level gain after 10 years 10 years | +30% | at which point GDP is almost 30% higher than it would have been without |
| The Software Industry and Artificial Intelligence: Nearly Universal Use, Evidence That Ranges from Large Gains to a Slowdown, and the Question of Who Gets Hired | Lab experiment: time to finish a programming task with an AI pair programmer Region and horizon not recorded | −55.8% | In a laboratory experiment, developers with an AI pair programmer completed a programming task 55.8% faster |
| The Software Industry and Artificial Intelligence: Nearly Universal Use, Evidence That Ranges from Large Gains to a Slowdown, and the Question of Who Gets Hired | Field experiments, 4,867 developers: completed tasks with a coding assistant Region and horizon not recorded | +26.1% | in three field experiments with 4,867 developers at three large companies, the number of completed tasks rose by 26.08% |
| The Software Industry and Artificial Intelligence: Nearly Universal Use, Evidence That Ranges from Large Gains to a Slowdown, and the Question of Who Gets Hired | RCT, 16 experienced open-source developers: completion time when AI tools allowed Region and horizon not recorded | +19% | allowing AI tools increased completion time by 19% |
| Legal Services and Artificial Intelligence: Real Productivity Gains, Fabricated Citations, and a Profession Learning to Verify | Law-student trial: productivity gains with a legal AI tool and a reasoning model Region and horizon not recorded | +90% (range 50 to 130) | a recent trial found productivity gains of 50% to 130% on five of six tasks |
| DEVELOPER PRODUCTIVITY IN THE ERA OF ARTIFICIAL INTELLIGENCE AND MICROSERVICES: A CASE STUDY OF GITHUB COPILOT INTEGRATION AT DUOLINGO | Duolingo, GitHub Copilot: engineering velocity in new or unfamiliar repositories Region and horizon not recorded | +25% | 25% among developers working in new or unfamiliar repositories |
| DEVELOPER PRODUCTIVITY IN THE ERA OF ARTIFICIAL INTELLIGENCE AND MICROSERVICES: A CASE STUDY OF GITHUB COPILOT INTEGRATION AT DUOLINGO | Duolingo, GitHub Copilot: engineering velocity in a familiar environment Region and horizon not recorded | +10% | a consistent 10% in a familiar environment |
| Longitudinal effects of ambient artificial intelligence scribe use on documentation burden and financial productivity in primary care: a quasi-experimental study | Primary care, AI scribe: note-writing time after 150 days Region and horizon not recorded | −16% (range −18 to −13) | increasing to 16% decrease by day 150 (IRR 0.84, 95% CI [0.82, 0.87]) |
| Longitudinal effects of ambient artificial intelligence scribe use on documentation burden and financial productivity in primary care: a quasi-experimental study | Primary care, AI scribe: work relative value units after 150 days Region and horizon not recorded | +2% | a 2% increase in wRVU (1.02, 95% CI [1.01, 1.04]) |
| Assessing the dynamic role of artificial intelligence in driving economic growth in sub-Saharan Africa | Share of the AI-growth effect that runs through industrial development, 36 countries, 2010-2022 Sub-Saharan Africa | 23.6% | approximately 23.6% of the total AI–growth effect operating indirectly through industrial development |
| Task Expansion with Generative AI: Experimental Evidence on Middle-Skilled, Early-Career Workers | Gain in task performance with AI on tasks beyond IT apprentices' training (percentage points) Germany | +28.5% (range 26 to 31) | productivity increases by 26–31 percentage points on tasks beyond apprentices’ formal training |
| Impact of Artificial Intelligence on Modern Software Development Lifecycle: Methodological Advancements, Developer Productivity, and Quality Engineering | Coding lead time with AI assistance Region and horizon not recorded | −55.8% | AI assistance compresses coding lead times by 55.8% |
| Impact of Artificial Intelligence on Modern Software Development Lifecycle: Methodological Advancements, Developer Productivity, and Quality Engineering | Code churn with ungoverned AI code generation Region and horizon not recorded | +19.4% | a 19.4% surge in code churn |
| Using artificial intelligence in insurance: ZhongAn Company as a case study | Employee productivity at insurer ZhongAn after AI automation (2019-2024) China | 30× | employee productivity has increased 30‑fold |
| Using artificial intelligence in insurance: ZhongAn Company as a case study | Claims settlement time at ZhongAn after AI automation China | −87.5% | the reduction in claims settlement time is 87.5% |
| AI Horizons: How the Organizations Seeing the Most AI Value Are Redesigning Work | Managers at large US firms saying AI is fully embedded in core processes US | 28% | Only 28% of managers and senior leaders say AI is fully embedded in their core business processes |