What did businesses say
before today’s AI conversation?
Earlier surveys recorded training, workflow changes, obstacles, and businesses selling AI-enabled products. Their questions help explain how the conversation developed.
Save PDF / Print opens your browser’s print options. Choose “Save as PDF” for a copy.
Read three evidence windows
The 2019 Annual Business Survey asks about technology during 2016–18. The first detailed BTOS AI supplement was collected December 2023–February 2024. The newer supplement was collected November 2025–February 2026. Each describes its own time, population, and AI definition.
These historical tables have now been reviewed and their original bytes preserved. They support a fuller account of what businesses reported. They do not establish an uninterrupted adoption series or an AI growth rate across the definition change.
2024: adoption involved people and processes
The earlier BTOS definition focused on AI used to produce goods or services. In the pooled historical supplement, 5.0% of in-scope businesses reported using AI in the preceding two weeks. This is the historical supplement estimate, rather than an October 2026 adoption rate.
In the last two weeks, did this business use Artificial Intelligence (AI) in producing goods or services? (Examples of AI: machine learning, natural language processing, virtual agents, voice recognition, etc.)
| Response | Estimate (%) | Standard error (pp) |
|---|---|---|
| Yes | 5.0 | 0.05 |
| No | 84.2 | 0.08 |
| Do not know | 10.8 | 0.08 |
Among businesses that had used AI in producing goods or services during the preceding six months, 20.8% reported training current staff, 19.7% developing new workflows, and 7.8% using vendors or consultants to install or integrate AI. These responses describe implementation actions; they do not measure spending or prove demand for a specific service.
In the last six months, to use Artificial Intelligence (AI), what changes did this business make?
| Response | Estimate (%) | Standard error (pp) |
|---|---|---|
| Trained current staff to use AI | 20.8 | 0.34 |
| Hired staff trained in AI | 2.0 | 0.10 |
| Purchased computing power or specialized equipment | 5.9 | 0.18 |
| Purchased cloud services or cloud storage | 12.4 | 0.26 |
| Changed data collection or data management practices | 8.0 | 0.21 |
| Developed new workflows | 19.7 | 0.40 |
| Used vendors or consulting services to install or integrate AI | 7.8 | 0.16 |
| Other | 7.6 | 0.13 |
| None | 50.5 | 0.40 |
In that AI-user group, 94.6% reported no change in total employment due to AI. A business-level response cannot be converted into a percentage of workers gaining or losing jobs.
In the last six months, how did the use of Artificial Intelligence affect this business’s total employment?
| Response | Estimate (%) | Standard error (pp) |
|---|---|---|
| Increased | 2.8 | 0.23 |
| Decreased | 2.6 | 0.13 |
| Did not change | 94.6 | 0.32 |
Among businesses not planning to use AI to produce goods or services, 80.9% said it was not applicable. Knowledge and privacy concerns also appeared. The respondent group differs from all businesses and from AI users.
Why does this business not plan to use Artificial Intelligence (AI) during the next six months in producing goods or services?
| Response | Estimate (%) | Standard error (pp) |
|---|---|---|
| Too expensive | 4.1 | 0.07 |
| AI is not a mature enough technology yet | 6.1 | 0.09 |
| Lack of knowledge on the capabilities of AI | 7.3 | 0.06 |
| Concerns about privacy/security | 6.6 | 0.05 |
| Concerns about bias | 2.8 | 0.04 |
| Lack of skilled workforce | 2.9 | 0.05 |
| Lack of required data | 2.2 | 0.03 |
| Laws and regulations prevent or restrict use of AI | 1.2 | 0.04 |
| Previous or current use of AI did not meet expectations | 0.9 | 0.02 |
| Other | 4.5 | 0.07 |
| AI is not applicable to this business | 80.9 | 0.14 |
Chamber interpretation
Training, workflow design, and integration were part of the business AI conversation before the newer survey definition. An entrepreneur can use these categories to investigate a customer problem and test an offer. The historical percentages do not forecast today’s revenue opportunity.
See all 15 questions in the earlier AI supplement
2024 question 1: In the last two weeks, did this business use Artificial Intelligence (AI) in producing goods or services? (Examples of AI: machine learning, natural language processing, virtual agents, voice recognition, etc.)
Respondent group: Statistics are representative of all companies in the target population.
2024 question 2: In the last six months, what types or applications of Artificial Intelligence (AI) did this business use in producing goods or services?
Respondent group: Statistics are representative of all companies in the target population.
2024 question 3: In the last six months, did this business use Artificial Intelligence to perform tasks previously done by employees in producing goods or services?
Respondent group: Statistics are representative of companies that reported using Artificial Intelligence to produce goods or services in the last six months.
2024 question 4: In the last six months, how many tasks previously done by employees were instead performed by Artificial Intelligence?
Respondent group: Statistics are representative of companies that reported using Artificial Intelligence to perform tasks currently done by employees in producing goods or services.
2024 question 5: In the last six months, did this business use Artificial Intelligence to perform operations previously performed by existing equipment or software in producing goods or services?
Respondent group: Statistics are representative of companies that reported using Artificial Intelligence to produce goods or services in the last six months.
2024 question 6: In the last six months, how did the use of Artificial Intelligence affect this business’s total employment?
Respondent group: Statistics are representative of companies that reported using Artificial Intelligence to produce goods or services in the last six months.
2024 question 7: In the last six months, to use Artificial Intelligence (AI), what changes did this business make?
Respondent group: Statistics are representative of companies that reported using Artificial Intelligence to produce goods or services in the last six months.
2024 question 8: During the next six months, do you think this business will be using Artificial Intelligence (AI) in producing goods or services? (Examples of AI: machine learning, natural language processing, virtual agents, voice recognition, etc.)
Respondent group: Statistics are representative of all companies in the target population.
2024 question 9: During the next six months, what types or applications of Artificial Intelligence (AI) do you think this business will use in producing goods or services?
Respondent group: Statistics are representative of companies that reported planning to use Artificial Intelligence to produce goods or services in the next six months.
2024 question 10: During the next six months, do you think this business will use Artificial Intelligence to perform tasks currently done by employees in producing goods or services?
Respondent group: Statistics are representative of companies that reported planning to use Artificial Intelligence to produce goods or services in the next six months.
2024 question 11: During the next six months, how many tasks currently done by employees will instead be performed by Artificial Intelligence?
Respondent group: Statistics are representative of companies that reported planning to use Artificial Intelligence to perform tasks currently done by employees in producing goods or services.
2024 question 12: During the next six months, do you think this business will use Artificial Intelligence to perform operations currently performed by existing equipment and software in producing goods or services?
Respondent group: Statistics are representative of companies that reported planning to use Artificial Intelligence to produce goods or services in the next six months.
2024 question 13: During the next six months, how do you think the use of Artificial Intelligence will affect this business’s total employment?
Respondent group: Statistics are representative of companies that reported planning to use Artificial Intelligence to produce goods or services in the next six months.
2024 question 14: During the next six months, to use Artificial Intelligence, what changes do you think this business will make?
Respondent group: Statistics are representative of companies that reported planning to use Artificial Intelligence to produce goods or services in the next six months.
2024 question 15: Why does this business not plan to use Artificial Intelligence (AI) during the next six months in producing goods or services?
Respondent group: Statistics are representative of companies that reported planning not to use Artificial Intelligence to produce goods or services in the next six months.
Original 2024 Census workbook ↗
March 28, 2024 publication announcement ↗ · Early-adoption research brief →
Earlier ABS: businesses also sold AI-enabled products
The older survey goes beyond buying or adopting technology. Table 114 asks whether companies sold AI or goods or services with AI during 2016–18. The national “yes” estimate is 0.4% of all in-scope companies. It is a historical result; it does not measure today’s producer population or the share of entrepreneurs succeeding in AI.
| Response | Estimate (%) | Source flag |
|---|---|---|
| Yes | 0.4 | — |
| No | 92.8 | — |
| Don't know | 6.8 | — |
Numerical standard errors are not supplied in the uploaded ABS table. “—” means no source flag; it does not mean zero uncertainty. Multiple-response items can overlap.
A different table asks companies producing AI or AI-enabled goods or services about their motivations. Among that producer group, 56.5% reported upgrading goods or services, 54.8% expanding their range, and 41.2% entering new markets or adapting existing products to new markets. Multiple motivations could be selected.
| Response | Estimate (%) | Source flag |
|---|---|---|
| To upgrade goods or services | 56.5 | — |
| To expand the range of goods or services | 54.8 | — |
| To enter new markets or adapt existing products to new markets | 41.2 | — |
| To increase or maintain market share | 36.1 | — |
| To adopt standards and accreditation | 16.5 | — |
| Some other reason | 14.8 | — |
Numerical standard errors are not supplied in the uploaded ABS table. “—” means no source flag; it does not mean zero uncertainty. Multiple-response items can overlap.
Chamber interpretation
Participation can include building or improving an offering, alongside using a tool internally. The survey gives us a documented producer perspective to bring into the Chamber conversation. It does not identify vendors, calculate market size, or show which business model will succeed today.
2016–18: AI as a production technology
Census conducted the 2019 ABS with NSF’s National Center for Science and Engineering Statistics. Its technology tables cover data year 2018 and refer to 2016–18. The reviewed NCSES release was published March 4, 2022. “2019” identifies the survey cycle, rather than the publication date.
The adoption table distinguishes testing from low, moderate, and high use in production or services. We retain these published categories instead of combining rounded percentages into a new headline rate.
| Response | Estimate (%) | Source flag |
|---|---|---|
| Did not use | 89.9 | — |
| Tested, but not used in production or service | 0.4 | — |
| Low use | 1.6 | — |
| Moderate use | 1.0 | — |
| High use | 0.4 | — |
| Don't know | 6.7 | — |
Numerical standard errors are not supplied in the uploaded ABS table. “—” means no source flag; it does not mean zero uncertainty. Multiple-response items can overlap.
Among companies using AI, 49.4% reported a motivation to improve quality or reliability, 37.1% to upgrade outdated processes, and 27.6% to automate tasks performed by labor. These are reasons for use, rather than verified improvements after implementation.
| Response | Estimate (%) | Source flag |
|---|---|---|
| To automate tasks performed by labor | 27.6 | — |
| To upgrade outdated processes or methods | 37.1 | — |
| To improve quality or reliability of processes or methods | 49.4 | — |
| To expand the range of goods or services | 23.9 | — |
| To adopt standards and accreditation | 11.8 | — |
| Some other reason | 18.2 | — |
Numerical standard errors are not supplied in the uploaded ABS table. “—” means no source flag; it does not mean zero uncertainty. Multiple-response items can overlap.
The preserved historical candidate decodes 13 AI-specific ABS tables: 84, 85, 94, 95, 96, 109, 114, 115, 124, 125, 126, 139, and 140. They cover adoption, motivations, workforce effects, barriers, industry, and company size. Three technical tables document the sample and response rates. The complete original table archives are also preserved.
Compare the questions before comparing the numbers
The earlier BTOS core question asked about AI in producing goods or services. Collection broadened to any business function in November 2025; the first publication using the new definition was December 4, 2025. A broader question can capture uses the earlier question missed.
Training, employment, and several barrier labels appear in both supplements, but the AI-user and non-planner groups also changed with the broader definition. Older technology-type questions differ from newer business-function questions. Employee generative-AI assistance and its task detail are new measures.
| Concept | 2024 | Newer | Comparison limit |
|---|---|---|---|
| Current AI use | 1 | 23 | Producing goods/services changed to any business function; retain separate rates. |
| Types and applications | 2 | 24 | Technology/application types changed to business functions; no itemwise trend. |
| Employee task replacement | 3 | 25 | Yes/no question changed to multi-select replacement/assistance/new tasks; AI-user population broadened. |
| Extent of task replacement | 4 | 26 | Small/moderate/large categories remain; qualifying AI-user population differs. |
| Software/equipment replacement | 5 | 27 | Older question restricts operations to production of goods/services. |
| Employment effects | 6 | 28 | Both ask total employment effects; qualifying AI use is broader in 2026. |
| Implementation changes | 7 | 29 | Training/workflows/cloud/data/vendor items overlap; purchase category also changed. |
| Expected AI use | 8 | 33 | Future adoption definition broadened; plans are not observed outcomes. |
| Expected applications | 9 | 34 | Types/applications changed to planned business functions. |
| Expected employee-task replacement | 10 | — | No matching expected-task-replacement question in the newer 13-question supplement. |
| Expected replacement extent | 11 | — | No matching expected-task-extent question. |
| Expected software/equipment replacement | 12 | — | No matching expected-equipment-replacement question. |
| Expected employment effect | 13 | — | New supplement measures past effect, not an identical AI employment forecast. |
| Expected organizational changes | 14 | — | New supplement measures past organizational changes; general AI-use expectations differ. |
| Reasons for no AI plans | 15 | 35 | Barrier labels overlap, but non-planners are defined using different AI scope. |
| Employee AI assistance | — | 30 | No direct counterpart in the older supplement. |
| Employee generative-AI assistance | — | 31 | 2024 large-language-model application item is not an equivalent employee-use measure. |
| Generative-AI work tasks | — | 32 | New employee-task detail is not comparable to older AI technology-type categories. |
ABS introduces further differences in sampling, reference periods, company scope, and nonresponse treatment. It uses 2017 NAICS. Counts underwent noise infusion and random rounding; missing answers for non-key categorical items were included in No/NA tabulations. Those methods belong with the figures.
ABS marker D means a value was withheld for disclosure protection; r means relative standard error above 50%. Neither is a zero. Quantitative standard errors are not supplied in these uploaded ABS tables. BTOS estimates retain their published standard errors and suppression markers.
Inspect the evidence behind this page
The downloads contain the reviewed national source rows, original answer labels, respondent groups, and worksheet coordinates. Each historical source stays separate from the newer national snapshot. These files contain public tabulations, rather than company microdata.