📊 Data & AI · Statistics hub · updated 2026-09-14
AI Analytics Statistics 2026
This page collects 53 statistics about AI and data analysis, and every one of them links to the organisation that published it and states the scope it was measured on. That last part matters more than the numbers themselves. The headline figures in circulation for "AI adoption" range from 4% to 78% — and the range is not a mistake. A 2024 official census of EU enterprises, a two-week recall survey of US businesses, a vendor-adjacent global survey and a controlled experiment are answering four different questions.
We built this page because most statistics roundups repeat figures without a traceable origin, which makes them useless in a report and dangerous in a decision. Sources here are statistical offices, academic papers, benchmark results and official corporate research. Where a widely quoted number could not be traced to its original publisher, it was left out — see the methodology note at the end for the specific example.
Cite this page
You are welcome to cite these figures in articles, reports and presentations. Each one links to its primary source, so you can cite that source directly, or cite this page for the collected set.
NoCodeCSV. "AI Analytics Statistics 2026." https://nocodecsv.com/ai-analytics-statistics (accessed 2026).
Figures were checked against the publishers' own pages on 2026-09-14. If you spot an error, tell us and it will be corrected with a note rather than silently removed.
Key takeaways
- 30.92% of EU enterprises (10+ employees) performed data analytics in 2025, up from 25.78% in 2023. Eurostat
- 20.0% of EU enterprises used AI in 2025 — up from 13.5% in 2024 and 7.7% in 2021. Eurostat
- Only 4.24% of EU enterprises used machine learning for data analysis — the narrowest official measure of AI-driven analysis. Eurostat
- 55.03% of large EU enterprises used AI in 2025, against 17.0% of small ones. Eurostat
- 18% of US firms used AI in a business function (32% employment-weighted), and 57% of adopters use it in three or fewer functions. US Census working paper CES-WP-26-25
- 76.56% of large EU firms analyse data, against 25.82% of firms with 10–49 employees: the size gap is the dominant pattern. Eurostat
- Official AI adoption is 20.0% (EU enterprises, 2025) and 18% of US firms (32% employment-weighted), while a self-reported global survey puts it at 78%. Scope, not disagreement, explains the gap. Eurostat / US Census / Stanford HAI
- 75% of knowledge workers use AI at work, but only 24% of leaders say AI is deployed organisation-wide. Microsoft Work Trend Index
- An AI pair programmer made developers 55.8% faster in a randomised controlled trial; a separate randomised trial at Google found about 21%. arXiv:2302.06590 / arXiv:2502.09479
- +15% issues resolved per hour across 5,172 support agents — but the most experienced workers saw small quality declines. arXiv:2307.01161
- Almost half of a data professional's time goes on loading and cleansing data — the traceable version of the widely quoted 80% claim. Anaconda, 2020 State of Data Science
- 80% of the global workforce says it lacks the time or energy to do its work — the binding constraint is attention, not tooling. Microsoft Work Trend Index 2025
- The best systems solve 16.33% of full BIRD tasks and reach 44.81% on the easier Lite split. BIRD benchmark
- ChatGPT's 40.08% execution accuracy against a 92.96% human result is the clearest argument for verifying AI-generated analysis. BIRD benchmark
- 46% of leaders already use agents to automate workflows, and 81% expect integration within 12–18 months. Microsoft Work Trend Index 2025
- 59% of leaders cannot quantify AI's productivity gains; 60% say their leadership lacks a plan. Microsoft Work Trend Index 2024
- USD 109.1 billion of private AI investment went into the United States in 2024 — about 12x China. Stanford HAI AI Index 2025
Adoption: who actually analyses data, and who uses AI to do it
How many companies perform data analytics at all?
About three in ten EU enterprises analysed data from at least one source in 2025 — and the share is climbing fast. This is the widest official measurement of analytics activity, and it is the base that every AI analytics figure sits on top of.
- 30.92% of EU enterprises with 10+ employees performed data analytics on data from at least one source (2025) Eurostat, Data analytics by size class of enterprise
- 25.78% in 2023 — a rise of 5.1 percentage points in two years Eurostat (same dataset)
How large is the gap between large and small companies?
The size gap is the single biggest pattern in every dataset on this page. Analytics is close to universal among large firms and still a minority activity among small ones, which is exactly why spreadsheet work dominates in SMEs.
- 76.56% of EU enterprises with 250+ employees perform data analytics, against 25.82% of those with 10–49 employees (2025) Eurostat
- Medium firms (50–249 employees) sit at 52.04% Eurostat
Do companies analyse more than one data source?
Most analytics is still single-source. Combining three or more sources — transaction data, customer data, web data, sensor data — is a minority practice even among large firms, which is the analytical gap that AI tooling is being sold into.
- 15.86% of EU enterprises analyse data from at least three sources (2025) Eurostat
- Among enterprises with 250+ employees the share rises to 50.64% Eurostat
How many companies use AI at all?
The official EU statistic is the hardest number on this page, because it comes from a survey of 157,000 enterprises run by national statistical authorities rather than a vendor poll. It also shows something the vendor numbers hide: adoption nearly tripled in four years, and it is still a minority activity.
- 20.0% of EU enterprises with 10 or more employees used AI technologies in 2025 — up 6.5 percentage points from 13.5% in 2024, and from 8.1% in 2023 and 7.7% in 2021 Eurostat, Use of AI in enterprises (2025 survey)
- 17.0% of small enterprises (10–49 employees), 30.36% of medium (50–249) and 55.03% of large enterprises (250+) used AI in 2025 Eurostat
- By sector: information and communication 62.52%, professional, scientific and technical activities 40.43%, construction 10.79% (the lowest of all sectors) Eurostat
- By country: Denmark 42.0%, Finland 37.8% and Sweden 35.0% at the top; Romania 5.2%, Poland 8.4% and Bulgaria 8.5% at the bottom Eurostat
Which companies are next — the ones that have not adopted yet?
Non-adoption is often 'not yet' rather than 'never'. The share of EU enterprises that had considered using AI but had not done so grew by 2 percentage points in a single year, and the figure is three times higher among large firms.
- 14.21% of EU enterprises that did not use AI had considered using it in 2025, up 2.02 percentage points on 2024 Eurostat
- Among large non-users the share considering AI was 36.54%, against 22.26% of medium and 12.65% of small enterprises Eurostat
How many companies use machine learning specifically for data analysis?
This is the narrowest and most honest number available from an official statistical office: machine learning applied to data analysis, as distinct from AI in general. Any vendor claiming a much larger number is measuring something else.
- 4.24% of EU enterprises used machine learning for data analysis (2024) Eurostat (indicator E_AI_TML)
What does the US government measure?
The US Census Bureau's Business Trends and Outlook Survey asks firms directly whether they used AI in the previous two weeks, which produces the most conservative adoption numbers in circulation — and the clearest picture of the size divide.
- 17–20% of US businesses reported using AI in the survey window US Census Bureau, Business Trends and Outlook Survey
- 37% of firms with 250+ employees, 32% of firms with 100–249 employees US Census Bureau
- Under 20% of firms with fewer than four employees US Census Bureau
How far does AI spread inside a company that does adopt it?
Adoption at the company level overstates what is actually happening inside it. A 2026 US Census working paper measured three layers separately — the firm, its business functions, and its workers' tasks — and found adoption is broad at the top and shallow underneath.
- 18% of US firms used AI in a business function (Nov 2025–Jan 2026), rising to 32% on an employment-weighted basis, with 22% expected within six months US Census Bureau working paper CES-WP-26-25, 'The Microstructure of AI Diffusion'
- Use reaches 50–60% (60–70% employment-weighted) among very large firms in Information, Professional Services and Finance US Census Bureau
- 57% of adopting firms use AI in three or fewer business functions; the most common are Sales and Marketing (52%), Strategy and Business Development (45%) and IT (41%) US Census Bureau
- Workers use AI in work-related tasks at 23% of firms (41% employment-weighted), and 65% of firms limit that use to three or fewer tasks US Census Bureau
- 66% of users rely on AI solely to augment tasks, and AI-related employment decreases were reported by only 2% of firms US Census Bureau
Why do published adoption numbers disagree so violently?
Because they measure different things. Placing the official-style numbers side by side is the fastest way to spot a statistic that has been inflated by its sampling frame — which is why this page states the scope of every figure rather than presenting one headline number.
- 20.0% — EU enterprises with 10+ employees, official survey, AI technologies used in business (Eurostat, 2025) Eurostat
- 18% of firms, 32% employment-weighted — US firms using AI in a business function (US Census working paper, Nov 2025–Jan 2026) US Census Bureau
- 17–20% — US businesses, official survey, AI used in the past two weeks (US Census BTOS, 2025–2026) US Census Bureau
- 78% — organisations self-reporting AI use in a global survey (Stanford HAI AI Index, 2025), up from 55% a year earlier Stanford HAI AI Index 2025
How many individual workers use AI?
Individual use runs far ahead of organisational deployment, which is the defining feature of the current phase: employees adopt AI on their own before their employer has a policy for it.
- 75% of global knowledge workers say they use AI at work; 46% of them started less than six months before the survey Microsoft Work Trend Index 2024
- 24% of leaders say their company has deployed AI organisation-wide, while 12% are still in pilot mode Microsoft Work Trend Index 2025
- 84% of developers are using or planning to use AI tools, and 51% of professional developers use them daily Stack Overflow Developer Survey 2025
Market and investment
How much money is flowing into AI?
Investment figures are the most reproducible part of the AI statistics landscape, because they come from financial databases rather than surveys. They also show how concentrated the money is in one market.
- USD 109.1 billion of private AI investment in the United States (2024) — roughly 12 times China (USD 9.3B) and 24 times the UK (USD 4.5B) Stanford HAI AI Index 2025
- USD 33.9 billion of global private investment into generative AI, up 18.7% year on year Stanford HAI AI Index 2025
Time and cost: what AI actually saves
How much faster do people work with AI assistance?
The strongest evidence comes from controlled experiments and enterprise telemetry rather than opinion surveys. The gains are real, but they cluster on repetitive, well-specified tasks — not on judgement.
- 55.8% faster task completion with an AI pair programmer in a randomised controlled trial Peng et al., 'The Impact of AI on Developer Productivity', arXiv:2302.06590
- +40.5% more pull requests completed in an engineer's heaviest AI-usage weeks, holding measured effort constant (observational dose–response study) arXiv, 'GitHub Copilot and Developer Productivity'
- Up to 50% time saved on documentation and autocompletion, 30–40% on repetitive coding, unit tests and debugging in an enterprise case study arXiv, enterprise evaluation of GitHub Copilot
- 33% of AI suggestions accepted, with 72% developer satisfaction, in a production deployment at ZoomInfo arXiv, 'Experience with GitHub Copilot for Developer Productivity at ZoomInfo'
- About 21% less time on a complex enterprise-grade task with AI assistance, in a randomised controlled trial with 96 full-time software engineers arXiv:2502.09479, randomised controlled trial
- +15% issues resolved per hour across 5,172 customer support agents — but the gains were uneven: less experienced workers improved in both speed and quality, while the most experienced saw small speed gains and small quality declines Brynjolfsson, Li & Raymond, 'Generative AI at Work', arXiv:2307.01161
How much of a data professional's time goes on preparing data?
This is the most misquoted statistic in the field. The often-repeated claim that analysts spend 80% of their time preparing data is normally published without a year or a source. The traceable version comes from the vendor survey that is usually being referred to, and its own wording is 'almost half'.
- 'Almost half of their time is spent on the combined tasks of data loading and cleansing' — survey of data professionals in 100+ countries Anaconda, 2020 State of Data Science
- Data visualisation takes about 21% of a data professional's time; modelling tasks, including selection, training, scoring and deployment, take the remaining third Anaconda, 2020 State of Data Science
What do users themselves report?
Self-reported savings are not evidence of productivity, but they explain adoption behaviour — and they are reported at levels no other enterprise software category reaches.
- 90% of AI users say AI saves them time; 85% say it helps them focus on important work Microsoft Work Trend Index 2024
- The heaviest 5% of Teams users summarised 8 hours of meetings with Copilot in a single month — the equivalent of one working day Microsoft Work Trend Index 2024
Is there time left to save?
This is the counterweight to every productivity claim. The constraint on analysis is not tool speed but available attention, and the workforce reports a severe shortage of it.
- 80% of the global workforce says they lack enough time or energy to do their work, while 53% of leaders say productivity must increase Microsoft Work Trend Index 2025
- 48% of employees and 52% of leaders describe their work as chaotic and fragmented Microsoft Work Trend Index 2025
Where AI analysis fails
How accurate are AI models at real analysis tasks?
This is the most useful block on the page for anyone deciding how much to trust an AI-generated number. On benchmarks built from real databases and messy schemas, the best systems still fail most of the time — and the gap to human performance is an order of magnitude in effort, not a rounding error.
- The best LLMs reached a 16.33% success rate on full BIRD tasks; on the easier Lite version, the leading model reached 44.81% BIRD benchmark (official site)
- o3-mini reached 24.4% success rate on c-Interact and Claude 3.7 Sonnet 17.78% on a-Interact in the interactive variants BIRD benchmark
- ChatGPT achieved 40.08% execution accuracy against a human result of 92.96% BIRD benchmark
- Vision-language models were 4–7x more likely to harmfully classify individuals with darker skin tones in the DABench auditing benchmark DABench, arXiv:2406.17974
What does that mean for a spreadsheet user?
It means the right workflow is verify-first: use AI to write the formula, generate the SQL, or describe the dataset, then check the result against a row count, a sanity total or the source file. An AI answer about your data is a draft, not a finding — and a wrong answer usually looks exactly like a right one.
- 40% execution accuracy, in practice, means a majority of generated queries are wrong on first attempt and must be checked BIRD benchmark
Where it is heading: agents and barriers
Are companies actually deploying AI agents?
Agents moved from demo to deployment in a single year, at least in the plans of the companies surveyed. The gap between the 46% automating workflows today and the 81% expecting integration within 18 months is where most of the 2026 procurement decisions sit.
- 46% of leaders say their company already uses agents to fully automate workflows or processes Microsoft Work Trend Index 2025
- 81% expect agents to be moderately or extensively integrated into their AI strategy within 12–18 months Microsoft Work Trend Index 2025
- 42% expect to be building multi-agent systems to automate complex tasks within five years Microsoft Work Trend Index 2025
What is blocking adoption?
Not price, and not capability. The barrier reported most often is the inability to prove that AI is working, followed by the absence of an internal plan — which is a measurement problem, and therefore a spreadsheet problem.
- 79% of leaders say their company needs to adopt AI to stay competitive, but 59% worry about quantifying the productivity gains, and 60% say leadership lacks a plan to implement it Microsoft Work Trend Index 2024
- 32% of managers plan to hire AI agent specialists within 12–18 months; 28% are considering an AI workforce manager role Microsoft Work Trend Index 2025
- Leaders at the fastest-moving firms use AI for data science work at 72%, against 55% for everyone else Microsoft Work Trend Index 2025
Methodology and scope notes
- Every figure is labelled with its scope. "EU enterprises with 10+ employees", "US businesses", "self-reported by leaders" and "global" are not interchangeable, and a number without its scope is not a statistic.
- Official statistics are quoted before surveys. Where a government or statistical office has measured the same thing, that measurement is listed first, even when it is the lower number.
- Contradictions are kept, not averaged. 4.24%, 13.48%, 17–20% and 78% all appear on this page because they measure different things. Averaging them would produce a number that is wrong about all four.
- Untraceable figures are excluded. The most repeated statistic in this field — that analysts spend 80% of their time preparing data — traces to a vendor survey whose own published wording is "almost half", and it is normally quoted without its year. The traceable version is published above with its source and date; the 80% figure is not. The same test is applied to any number whose only online presence is other blogs.
- Benchmarks are stated as published. BIRD and DABench results change as models improve, so both the benchmark name and its split (full or Lite) are given, and the publisher's own page is linked.
Frequently asked questions
How many companies use AI for data analysis?
On the narrowest official measure, 4.24% of EU enterprises used machine learning for data analysis in 2024, while 20.0% of EU enterprises used AI technologies of any kind in 2025 (Eurostat). On the widest definition — any analytics performed on any data source — 30.92% of EU enterprises did data analytics in 2025. In the United States, 18% of firms used AI in a business function (32% on an employment-weighted basis) in the 2026 Census survey, while a global self-reported survey puts organisational AI use at 78% (Stanford HAI AI Index 2025). The range reflects scope, not disagreement.
What percentage of businesses use AI?
Between 17% and 20% on official measurement: 20.0% of EU enterprises with 10 or more employees (Eurostat, 2025, up from 13.5% in 2024) and 17–20% of US businesses (US Census Bureau Business Trends and Outlook Survey, 2025–2026). The 2026 Census AI supplement reports 18% of US firms using AI in a business function, rising to 32% when weighted by employment. Large firms are three to four times more likely to use AI than the smallest ones.
How much time does AI actually save?
Reported savings are large but come from two different kinds of evidence. Controlled experiments show 55.8% faster completion on a well-specified coding task, and enterprise telemetry shows 30–50% savings on documentation and repetitive work. Self-reports are higher: 90% of AI users say it saves them time, and the heaviest Copilot users in Teams summarised eight hours of meetings in a month.
How accurate is AI at analysing data?
On BIRD, a benchmark built from real databases, the best models reached a 16.33% success rate on full tasks (44.81% on the easier Lite split), and ChatGPT reached 40.08% execution accuracy against a human result of 92.96%. Accuracy drops further when the data is non-English, the schema is undocumented, or the question is ambiguous — so an AI answer about your data should be verified before it is used.
How many companies are using AI agents?
46% of surveyed leaders say their company already uses agents to fully automate workflows or processes, 81% expect agents to be moderately or extensively integrated within 12–18 months, and 42% expect to run multi-agent systems within five years (Microsoft Work Trend Index 2025).
Can I cite this page?
Yes — that is what it is for. Every figure links to the publisher's own page and states its scope and year. A suggested citation is in the box above; a link back to this page is appreciated and no permission is required.
Now Check AI Against Your Own Data
Upload a CSV and ask for the totals, the trend or the odd rows — then compare the answer with a row count, because 40% execution accuracy is what the benchmark says.
Related reading
- Analysing survey data from a CSV with AI— a worked example of the verify-first workflow this page argues for.
- Excel row limit: why your CSV will not fit— the size constraint that sends analysis work to scripts in the first place.
- Count rows in a CSV file— the single check that catches most AI analysis errors.
- Convert text to CSV— getting a messy export into a shape a tool can read.
- CSV analyserand Excel data analysis — the tools these statistics are about.
Compiled by the NoCodeCSV team from primary sources, 2026-09-14. Scope and year are stated for every figure so that a reader can judge whether two numbers are comparable.