Official statistics, central-bank surveys and consultancy research on AI adoption, agents, spend, value and jobs, reconciled, with three NKK models for reading the numbers.
Written for: CIOs, heads of AI, strategy and transformation leads, board members
14 chapters, 57 cited sources, 27 models and charts. The first 2 chapters are open; the rest unlock with the form.
of US employer businesses used AI in any function in the prior two weeks (21 Sep-4 Oct 2026)
The most quoted number in enterprise technology, the share of companies using AI, has no single value. Official statistics put it at 23.8% of US employer businesses (prior two weeks, 21 September to 4 October 2026) [1] and 20.0% of EU enterprises in 2025 [2]. Executive panels report 69% across the US, UK, Germany and Australia [3], and the ECB's scale, which includes experimental use, gives about 70% in the euro area [4]. Among employed US adults, 45.2% use generative AI for work [5]. These answer different questions; they are not rival estimates of one quantity.
Our thesis is that adoption alone is a poor basis for management decisions and for comparing sources. Breadth grew by about 6.5 to 7 points over the latest reading period in each US, EU and Canadian series, partly on wider questionnaires [1][2][6], but depth is thin: 7% of euro area firms call their use significant [4]. Value claims swing from 98% positive ROI among leaders whose organisations are investing in AI [7] to 5% value at scale in BCG's 2025 study [8]; BCG's September 2026 index reports 7.5% future-built plus 41% scaling [9]. Measured effects on productivity and jobs are small and early, except hiring of young workers in exposed occupations [10]. Agent adoption runs from single digits to 91% because surveys count different things [11][12].
This report is a synthesis of public research; it does not contain primary survey data. It offers three NKK models for reading the numbers.
of US employer businesses used AI in any function in the prior two weeks (21 Sep-4 Oct 2026)
Evidence is weighted in four layers. Official statistics from the US Census Bureau [13], Eurostat [14], the UK Office for National Statistics [15], Statistics Canada [6] and the OECD [16] rest on large samples of firms. Central-bank and academic research from the ECB [4], the Federal Reserve [17][18], the BIS [19] and NBER [3] adds executive panels and linked records. Consultancy and vendor surveys from BCG [8], Deloitte [20], KPMG [21], EY [7], Gartner [22] and Microsoft with LinkedIn [23] are timelier and mostly reach senior executives, often at large firms, but they are self-reported, so we give sample size and population each time. The Stanford AI Index [11] is a compilation; its economy chapter does not name the survey behind its 88% adoption figure, so we treat that as a pointer, not a measurement.
Every number traces to a page we opened. NKK models are marked as ours, and their scores and bands as judgement.
The evidence stops on 13 October 2026. The Census Bureau and the ONS publish next on 22 October 2026 [1][24], and Eurostat plans an update in December 2026 [14]. BCG's September 2026 Applied AI Index is cited from its press release, not the full report [9]. The OECD's January 2026 announcement could not be opened directly, so its firm-level figures come from a European Commission page that reproduces it [25]. McKinsey, PwC, the World Economic Forum and MIT NANDA are widely quoted but not cited, because we could not open their primary pages.
Every figure and regulatory statement in this piece traces to a numbered source below. Links open the original publisher in a new tab.
The published failure rates for AI and agent projects measure different things. Here is what the evidence supports, why pilots stall, and a six-gate scorecard for deciding what earns production funding.
Self-reported benefits are common, yet P&L-visible value is a minority. Evidence on what AI delivers, why gains stall, and three NKK models for conversion, baselines and attribution.
Dated, source-checked status of AI regulation across the EU, US, UK, India, China and Asia-Pacific as of 13 October 2026, with three NKK models for managing regulatory change.
