Two numbers that are not measuring the same thing

Nearly a fifth of GDP growth sounds like an economy being remade by AI. The Bank of England's calculation is narrower. Staff combined output from publishing, computer programming and consultancy, and information services, then adjusted for those sectors' longer-run growth trend. The resulting proxy suggests that novel AI-related activity accounts for just under a fifth of UK GDP growth since early 2026. The Bank acknowledges that these industries also produce goods and services unrelated to AI, making the estimate an approximation rather than a direct measurement.

Bailey made the same point in person, in more cautious language. The activity, he said, "is concentrated in a subset of sectors," and the Bank's Agents report that the strength "has not been broad based." The report's verdict on the mechanism is one sentence long: the contribution "appears to reflect activity in the upstream development and implementation of AI technologies rather than broader gains from their adoption elsewhere in the economy."

What using AI has done, on firms' own account

Firms responding to the Bank's Decision Maker Panel survey estimate that AI has raised their productivity level by around 0.5% cumulatively over the past three years, in line with similar impacts reported in other countries. The Bank interprets that as a contribution of approximately 0.1–0.2 percentage points to annual productivity growth since early 2024. The Bank treats the wider pickup in measured productivity since 2024 as mostly cyclical, and its sector evidence cuts against a simple AI story. Productivity has risen in wholesale, retail, accommodation and food services, sectors with little AI exposure, as well as in information and communication. And in the AI-exposed sectors, it has not risen above its pre-Covid average.

None of this proves AI is doing nothing inside firms. A survey estimate is a blunt instrument, three years is early, and the Bank says so itself. What it does establish is the current distance between the two claims usually bundled into one headline. AI-related development and implementation are contributing visibly to measured economic activity. Evidence that AI adoption is raising productivity across the wider economy remains much smaller. The two estimates come from different methods and cannot be subtracted from one another.

Fewer hires, not more firings

The employment evidence is similarly specific, and worth separating from the predictions around it. Firms responding to the DMP survey report close to zero aggregate employment impact from AI so far. Separately, the Bank's regional Agents hear that AI adoption is reducing the likelihood of hiring or replacing departing workers, without materially increasing the likelihood of dismissals. Neither finding rules out job losses at individual companies.

That distinction matters for the forecast firms gave the Bank. Firms expect AI to contribute about 0.9 percentage points a year to productivity growth over the next three years, compared with the Bank's estimate of 0.1–0.2 percentage points annually since early 2024. That would be roughly 4.5 to nine times the historical contribution, although the comparison combines a backward-looking estimate with firms' expectations. Respondents expect the projected productivity improvement to be split roughly equally between faster output growth and slower employment growth. That is a forecast about future output and staffing, not evidence that AI has already caused large-scale layoffs. Read together, the survey describes a bet that the hiring channel, already visible, widens considerably. The Bank's central projection does not underwrite the bet: productivity growth there falls back toward its assumed 1% trend, with AI listed as an upside risk whose size and timing remain, in the Bank's words, subject to large uncertainty.

Sources

  1. [1] Bank of England, "Monetary Policy Report — November 2026," published 5 November 2026, read in full this run: Box C for the AI activity proxy (its SIC 58, 62 and 63 composition, the trend adjustment, the "just under a fifth of GDP growth since the start of 2026" estimate, the upstream development and implementation finding, and the demand-boost assessment); Box D for the Decision Maker Panel estimate of about 0.5% productivity gain over three years (0.1–0.2 percentage points a year since early 2024), the sector evidence including AI-exposed sectors not exceeding pre-Covid averages, the cyclical explanations, the close-to-zero aggregate employment impact, the Agents' hiring-versus-dismissal finding, and firms' expectation of about 0.9 percentage points a year over the next three years; Section 1.1 for AI-related product demand and memory chip prices adding to import price pressureRead source
  2. [2] Bank of England, "Monetary Policy Report press conference: Opening remarks by Andrew Bailey, Governor," 5 November 2026, read in full this run: "initial signs that growth in some sectors of the UK economy is being driven by AI-related investment," the AI Activity Proxy and the "just under a fifth" figure in the Governor's own summary, the passage stating the activity "is concentrated in a subset of sectors" and that Agents' intelligence suggests strength "has not been broad based," and the raising of Bank Rate by 25 basis points to 4%Read source
  3. [3] Secondary coverage, for timing only: The Register, "Bank of England pours cold water on AI productivity boom," 6 November 2026, and AI Weekly's summary published 7 November 2026. No load-bearing claim in this article rests on either.