Nshipyard Canada · The Adoption Inversion

On the one AI question both countries asked, Canada was ahead.

Statistics Canada asked businesses whether they used AI to produce goods or deliver services. The U.S. Census Bureau asked businesses nearly the same question. In 2025 the answers were 12.2% in Canada and 10.0% in the United States. This page charts the two official series, documents exactly how they were harmonized, and marks where the comparison stops being honest.

See the chart Read the methodology

12.2%

of Canadian businesses used AI to produce goods or deliver services, Q2 2025 (StatCan), up from 6.1% a year earlier

10.0%

of U.S. businesses used AI in producing goods or services, September 2025 (Census BTOS, legacy wording)

19.2%

Canada's latest reading, Q2 2026, the most recent StatCan release at the time this page was built

Nov 2025

the U.S. Census changed the question to broader wording; everything after that date is a different series and is not charted here

The chart

Same question, two governments.

Share of businesses using AI, Canada (red) vs United States (grey). Canada's points come from the Canadian Survey on Business Conditions, asked once a year. The U.S. line is the Census Bureau's Business Trends and Outlook Survey legacy question, asked every two weeks from September 2023 to November 2025.

Canada (StatCan) United States (Census, legacy wording)

Canada's points are annual (Q2 2024, Q2 2025, Q2 2026). The U.S. series is biweekly; the labeled points are the September fortnights.

Same question, both governments. Canada was ahead.

Sectors

Where the lead is widest.

2025, same three sectors on both sides of the border. Canada's figures are Q2 2025; the U.S. figures are the September 2025 fortnight of the legacy series.

CanadaUnited States

Information and cultural industries / Information

StatCan groups cultural industries with information; the U.S. NAICS 51 is information only.

35.6%

29.3%

Professional, scientific and technical services

31.7%

23.9%

Finance and insurance

30.6%

17.1%

Sector mapping is documented in the methodology. No adjustment was made for the grouping difference.

Firm size

Adoption by firm size, U.S. only.

September 2025 fortnight, Census BTOS legacy question. Adoption is U-shaped: the smallest and the largest firms report the most AI use.

U.S. only. Statistics Canada publishes firm-size cuts in Table 33-10-1005, which was not retrievable when this page was built; the Canadian panel will be added in the next update.

Caveats

What this does not show.

01
The reference windows differ. Statistics Canada asks about AI use over the 12 months preceding the survey; the Census Bureau asks about the last two weeks. A 12-month window mechanically raises the share answering yes, so part of Canada's lead reflects the window, not just adoption. This is the largest honest discount on the headline.
02
The U.S. changed the question on November 17, 2025, from 'in producing goods or services' to 'in any of its business functions'. The Census Bureau started a new time series on December 4, 2025. Only the pre-break fortnights are charted here; nothing after the break is comparable.
03
U.S. readings published after the question change, including figures circulating in curated datasets, were not verified against Census files at build time and are not charted. Do not splice them onto this line.
04
Both surveys weight by firm count, not by employment. In Canada, small and medium businesses employ about 64% of private-sector workers (ISED, Key Small Business Statistics 2024). Economies with more small firms mechanically show lower firm-count-weighted adoption at identical per-worker use.
05
The curated dataset that first popularized the U.S. series explicitly warns its headline figures are not comparable across its own cuts and does not cover Canada. Its numbers were not used here; every U.S. figure on this page comes from the Census Bureau's own files.
06
Canada's firm-size cuts (StatCan Table 33-10-1005) could not be retrieved at build time, so the firm-size panel is U.S.-only. The sector panel uses the three sectors published on both sides.

Why this exists

The claim being rebutted.

Through 2026, commentary has repeated that Canadian businesses trail U.S. businesses in AI adoption, including a Macdonald-Laurier Institute piece in October 2026 and widely shared KPMG-sourced claims of 72% U.S. versus 35% Canadian large-firm use. Those comparisons mix different questions, different reference windows, and different firm-size cuts. On the one question both governments asked in nearly identical words, the official numbers say the opposite. That is the entire argument of this page, and the caveats above are part of it.

Methodology

How the comparison was built, and where it is weak.

01
Canada: Canadian Survey on Business Conditions, Q2 2025. 12.2% of businesses reported using AI to produce goods or deliver services over the 12 months preceding the survey, up from 6.1% in Q2 2024. Source: Statistics Canada analytical publication 11-621-m2025008, released 2025-06-16, which cites CSBC Table 33-10-1004-01. Q2 2026: 19.2% (publication 11-621-m2026010). Sample: 21,357 businesses invited, 9,103 respondents. Collection: April 1 to May 5, 2025.
02
United States: Census Bureau Business Trends and Outlook Survey, legacy question 'In the last two weeks, did this business use Artificial Intelligence (AI) in producing goods or services?' National 'Yes' shares from the Census Bureau's own AI Core Questions file (original wording series), pulled 2026-10-09. September reference points: 3.7% (Sept 2023), 6.0% (Sept 2024), 10.0% (Sept 2025). The full biweekly series of 54 fortnights is charted and downloadable.
03
Question wording is nearly identical in verbs and scope; the reference window is not (12 months vs two weeks). The window difference is disclosed as the largest caveat, not adjusted away.
04
Wording break: the Census Bureau changed the question on November 17, 2025 and began a new time series with data released December 4, 2025 (Census notice, December 3, 2025). Only fortnights before the break are used. Post-break values are not charted because they were not verified against Census files.
05
Sector mapping: StatCan 'information and cultural industries' to BTOS NAICS 51 (Information); 'finance and insurance' to NAICS 52; 'professional, scientific and technical services' to NAICS 54. StatCan's grouping includes cultural industries alongside information; the U.S. 51 is information only. Documented, not adjusted.
06
Table-number note: the StatCan publication cites Table 33-10-1004-01. The StatCan bulk CSV download served stale content for this cube at build time, so headline and sector figures are transcribed from the publication text, which is the primary source. The transcription was checked twice against the publication.
07
Quarterly update procedure: each quarter, re-pull the CSBC AI analysis (catalogue 11-621-m) and the Census BTOS AI Core Questions file; append the new headline point and September-equivalent U.S. point; re-verify the September labels against the file; if the Census changes wording again, freeze the series and draw a new break line.

For developers

Query it from code, or from an agent.

Three consumption paths, same canonical data. REST for applications, OpenAPI for integration, MCP tools over streamable HTTP for AI agents.

Endpoints

GET

/api/v1/series

Harmonized Canada vs US AI-adoption series: StatCan Q2 points plus the full 54-fortnight Census BTOS legacy series

{
  "canada": [ { "quarter": "Q2 2025", "pct": 12.2 }, … ],
  "us_sept_points": [
    { "label": "Sept 2023", "pct": 3.7 },
    { "label": "Sept 2024", "pct": 6.0 },
    { "label": "Sept 2025", "pct": 10.0 } ],
  "us_fortnightly": [ { "fortnight": "202319",
    "ref_start": "08/28/2023", "pct": 3.7 }, … ]
}

Try it →

GET

/api/v1/sectors

2025 sector comparison: three sectors on both sides (Canada Q2 2025 vs US Sept 2025 fortnight)

{
  "sectors": [ {
    "sector": "Finance and insurance",
    "canada_pct": 30.6, "us_pct": 17.1,
    "mapping_note": "StatCan industry grouping to BTOS NAICS 52"
  }, … ]
}

Try it →

GET

/api/v1/meta

Headline comparison, key caveat, sources, and pull date

{
  "headline_2025": { "canada_pct": 12.2,
    "canada_label": "Q2 2025",
    "us_pct": 10.0, "us_label": "Sept 2025" },
  "key_caveat": "Reference windows differ: 12 months vs two weeks…"
}

Try it →

OpenAPI spec

GET /api/openapi.json

Connect your agent

Put this data to work inside your AI tools.

Pick your harness, copy the prompt, send it to your agent. Your agent runs the setup itself.

Copy and send this to Claude Code

Set up the Canada AI Adoption Inversion MCP server so I can query it from here.
1. Run: claude mcp add --transport http canada-ai-adoption-inversion https://this-site.example/mcp
2. Run `claude mcp list` to confirm it connected.
3. Show me the harmonized Canada vs US AI adoption series and the sector comparison, and show me the result.

Data