Most 2025 AI predictions can be judged only against a specific measure, and the measure often changes the answer. Gartner forecast in July 2024 that at least 30 percent of generative AI projects would be abandoned after proof of concept by the end of 2025, and by 2026 it was saying the real share was at least 50 percent. In Statistics Canada's spring 2026 survey, 19.2 percent of Canadian firms reported AI use, a rise from 12.2 percent, while the Stanford AI Index 2026 put organizational adoption at 88 percent. Those figures count different populations, so none of them confirms or refutes the others.
Key takeaways
- A prediction is checkable only if it names a population, a measure, and a date; "AI will be everywhere" names none of them.
- Gartner's July 2024 forecast of at least 30 percent abandonment was low by its own later account: a 2026 Gartner article says at least 50 percent of generative AI projects were abandoned after proof of concept by the end of 2025.
- Statistics Canada reported that 19.2 percent of businesses reported AI use for goods or services in the second quarter of 2026, up from 12.2 percent in 2025 and 6.1 percent in 2024.
- The Stanford AI Index 2026 reports 88 percent organizational adoption, up from the 78 percent in the 2025 edition; the gap with the Canadian figure comes mainly from who was asked and what counted as use.
- Gartner's next dated forecast, that over 40 percent of agentic AI projects will be canceled by the end of 2027, cannot be scored until 2028.
What makes a prediction checkable
A year-ahead forecast helps a reader only if someone can later mark it right or wrong. That takes three things: who is counted, what is measured, and by when. "30 percent of generative AI projects will be abandoned after proof of concept by the end of 2025" meets the test on paper. "AI will transform work" does not, because no figure published in 2026 could contradict it.
Even a checkable forecast can be hard to settle. The forecaster may never publish a follow-up, and independent surveys often count something else. The cases below compare three kinds of 2025-era claims with data released since.
A project failure forecast and its outcome
Gartner issued its press release on July 29, 2024, ahead of its Data and Analytics Summit in Sydney. Rita Sallam, a distinguished vice president analyst at the firm, said executives were impatient for returns on generative AI. Gartner named poor data quality, inadequate risk controls, escalating costs, and unclear business value as the reasons projects would stall.
The outcome was worse than the forecast. A later Gartner article titled "Why Half of GenAI Projects Fail" states that by the end of 2025 at least 50 percent of generative AI projects had been abandoned after the proof-of-concept stage, citing the same four causes. Gartner repeated the figure on social media in 2026. The article page returned an access error to scripted requests, so the wording comes from search-result text and should be checked on Gartner's site. A separate survey by S&P Global Market Intelligence, as reported by CIO Dive, found 42 percent of companies abandoning most of their AI initiatives, up from 17 percent the year before, in a sample of about 1,000 respondents. That measure counts companies and "most initiatives," so it points the same way without testing the same claim.
Adoption forecasts and the definition problem
The Stanford AI Index 2026 reports organizational adoption of AI at 88 percent. The 2025 edition had reported 78 percent for 2024, up from 55 percent the year before. Statistics Canada, in an analysis released June 11, 2026 from the Canadian Survey on Business Conditions, put AI use at 19.2 percent of businesses in the second quarter of 2026. The 2025 reading was 12.2 percent, and the 2024 reading 6.1 percent. Collection ran from April 1 to May 6, 2026, and respondents described the previous 12 months. Data analytics led the list of uses among adopters.
A reader comparing 88 and 19.2 might conclude that Canada lags badly. Neither source supports that. The Canadian question asks about a specific use in producing goods or delivering services, and the sample covers businesses of every size across the national economy. The Stanford figure comes from a survey of organizations that the index page does not describe in the same terms. Respondents, size mix, and wording can produce a large gap before any real difference in practice is considered.
Forward-looking numbers exist too. In the third-quarter 2026 survey, released August 31, 2026, 25.2 percent of Canadian businesses said they intend to adopt AI within a year, against 14.5 percent in the prior survey, and 52.7 percent had no plans. Among those with no plans, 79.1 percent said AI was not relevant to their goods or services.
On jobs, the 2025 survey found most adopters (89.4 percent) reporting no change in employment, 4.3 percent an increase, and 6.3 percent a decrease. The 2026 analysis does not repeat those shares and calls the employment effect unclear, so the 2025 split is the latest figure this page can cite. It does report that 44.4 percent of adopting firms changed their training or staffing practices because of AI.
Trend measures that are easier to check
Some claims concern quantities that can be tracked continuously. The AI Index 2025 reported that training compute for notable models doubles roughly every five months and that the cost of querying a GPT-3.5-level system fell more than 280-fold between November 2022 and October 2024. The 2026 edition adds that industry produced over 90 percent of notable frontier models in 2025 and that documented AI incidents rose to 362 from 233 in 2024.
These are measurements of the past. A prediction that extends a measured trend, such as continued cost declines, is easier to evaluate than one that depends on adoption choices or policy. Trends can still break, so a measured doubling period describes what happened and promises nothing.
Forecasts and what the data showed by October 2026
| Claim or measure | Source | Figure | Main caveat |
|---|---|---|---|
| Generative AI projects abandoned after proof of concept by end of 2025 | Gartner, July 29, 2024 (forecast); Gartner article, 2026 (outcome) | At least 30 percent forecast; at least 50 percent reported | Outcome wording not opened directly |
| Canadian businesses using AI to produce goods or deliver services | Statistics Canada, June 11, 2026 | 19.2 percent, up from 12.2 percent | Narrow wording, 12-month look-back |
| Organizations using AI | Stanford AI Index 2026 | 88 percent, up from 78 percent in the 2025 edition | Different survey and definition |
| Agentic AI projects canceled by end of 2027 | Gartner, June 25, 2025 | Over 40 percent (forecast) | Cannot be scored before 2028 |
Questions to ask before trusting a forecast
- Which group is counted: all firms, large firms, or survey respondents who chose to answer?
- What counts as use, and over what time window?
- Did the forecaster publish a later result, or only the prediction?
- Is the source a statistical agency, a research institute, or a seller of related services?
- Does the number measure something that happened or something expected?
FAQ
Why do AI adoption percentages differ so much between reports?
Surveys differ in who they ask, how large the organizations are, and how use is defined. A national statistics agency surveying all businesses usually reports a lower share than a poll of organizations already engaged with AI.
Did the Gartner forecast come true?
The direction held and the size did not. Gartner forecast at least 30 percent abandonment after proof of concept by the end of 2025, and its own later article reports at least 50 percent.
Are predictions about jobs checkable yet?
Partly. Employment changes reported by AI-using Canadian businesses were small in 2025, and Statistics Canada called the 2026 effect unclear. A year or two of survey responses cannot confirm or rule out longer-term effects.
Where can readers find official AI use figures?
Statistics Canada publishes Canadian business survey results, and the Stanford Institute for Human-Centered Artificial Intelligence publishes the AI Index each year. Each states its own method.
Related reading
- AI development and adoption by country
- Where AI came from
- AI compared with superintelligence
- A beginner learning path for AI
More in AI Basics.
Last verified: October 2026
Sources
- Analysis on artificial intelligence use by businesses in Canada, second quarter of 2026, Statistics Canada, June 11, 2026
- The Daily, Canadian Survey on Business Conditions, third quarter 2026, Statistics Canada, August 31, 2026
- Use of artificial intelligence by businesses in Canada, Statistics Canada, June 16, 2025
- The AI Index Report 2026, Stanford Institute for Human-Centered Artificial Intelligence, 2026
- The AI Index Report 2025, Stanford Institute for Human-Centered Artificial Intelligence, 2025
- Gartner Predicts 30% of Generative AI Projects Will Be Abandoned After Proof of Concept By End of 2025, Gartner, July 29, 2024
- Why Half of GenAI Projects Fail, Gartner, 2026
- Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027, Gartner, June 25, 2025
- AI project failure data, S&P Global Market Intelligence survey, as reported by CIO Dive, 2025