Evidence
Every figure on this site names its source, its publication and its date. Figures more than three years old are labelled with their age. Superseded figures get replaced and the replacement is noted.
The four rules
- Every figure names its source, its publication and its date.
- Figures more than three years old carry their age.
- Superseded figures are replaced, and the replacement is noted.
- Figures I cannot trace to a primary source are not used, and the notable ones are listed below with the reason.
Rule four is the one nobody else publishes.
Always state the base
“35% of UK businesses with ten or more employees” is honest. “35% of UK businesses” is not. Micro businesses adopt less and dominate the business count, so leaving them out changes the number a lot. The apparent conflict between 25% and 35% in circulation is a base difference, not a contradiction.
The two figures that both read 11%
There are two live statistics that both read 11% and mean different things, and they never appear in the same sentence. Both are linked below.
- ONS, June 2026: 11% of businesses with 10 or more employees report that more than half their workforce has received AI-related training.
- DSIT and Ipsos, fieldwork spring 2024: 11% of employers say they or their employees undertook any AI training in the last 12 months.
Numbers I do not use, and why
| Claim in circulation | Why it is not here |
|---|---|
| “60% of employers cite AI skill gaps” | This figure does not appear in the DSIT and DCMS employer survey it gets attributed to. The report's actual skills-gap figures are 25% and 27%, on a base of 373 rather than all employers. The likely origin is a separate line about six in ten employers not using AI and not planning to, which is a non-adoption figure, not a skills figure |
| “11% of employers provided AI training” | The source says employers “have undertaken some form of AI training”, which includes self-directed training. “Provided” overstates it |
| “50% report uncertainty about what training to buy” | The source says “unsure what training is relevant for the business”. Given I sell training, that gloss is one a sceptical reader would rightly catch |
| “25% of UK businesses use AI” | Correct for December 2025, all size bands. Superseded by the June 2026 figures |
| “15% planning to adopt AI in the next 12 months” | The source says the next three months |
The employer survey, and its date
The DSIT and DCMS employer survey was published on 28 January 2026. Its fieldwork ran from 19 March to 7 June 2024. That is a two-year lag in the fastest-moving area on this site. It gets cited as spring 2024 fieldwork, or not at all.
What I do not know
- There is no official UK measure of adults' AI skills.
- Nobody has good evidence that digital skills training works at scale.
- There is no citable UK benchmark day rate for adult digital-skills training sold to employers, councils and housing associations. My prices are hypotheses, and they are on the organisations page as such.