Summary

• DWP recorded 128,779 AI-related course completions between 1 May 2025 and 30 April 2026 at a reported cost of £112,995.
• Two AI for All modules accounted for 126,046 completions, while 1,620 related to Senior Civil Service training.
• The figures count course completions rather than unique employees and cannot be treated as a workforce headcount.

The Department for Work and Pensions recorded 128,779 completions of AI-related training over 12 months, although almost all of the activity came through two introductory modules rather than specialist technical programmes.

The figures cover learning completed between 1 May 2025 and 30 April 2026 and were obtained through a Freedom of Information request analysed by the Parliament Street think tank. DWP reported £112,995 of expenditure on the AI-related training covered by the response.

The largest component was the AI for All pathway. Part One accounted for 73,045 completions and Part Two for 53,001, producing a combined 126,046. Senior Civil Service training accounted for another 1,620, leaving 1,113 completions spread across nine more specialised programmes.

Almost 98% of the total therefore came from the two foundation modules. Introductory AI literacy can sensibly be deployed across a large organisation, while advanced machine-learning and specialist programmes apply to a much narrower workforce, so the distribution describes a broad literacy programme with a smaller technical component.

Completions do not equal employees

The dataset counts courses completed rather than unique members of staff. Someone who finished both parts of AI for All contributes twice to the headline total, while employees taking additional programmes can appear several times.

The figures therefore do not show that 128,779 individual DWP employees received AI training. They measure learning activity and cannot establish the number of distinct staff involved, the depth of their capability or whether the courses changed how participants used AI at work.

That distinction becomes important as government departments use training volumes to demonstrate readiness for wider AI adoption. Completion figures show reach, but operational capability depends on what people learned, whether they can apply it appropriately and whether the systems around them allow AI to be used safely.

Foundational literacy still has a significant role because public sector adoption extends far beyond specialist developers. Employees using general-purpose tools need to understand data handling, unreliable outputs, privacy and the need for oversight even when they never build a model themselves.

The government has placed that broader objective inside its AI Skills Boost programme, which aims to equip 10 million UK workers with practical AI skills by 2030. DWP is among the public sector partners attached to the programme, which includes free foundation training benchmarked through Skills England.

Training is easier to measure than capability

Foundation skills cover practical use alongside responsible operation, including how to instruct AI tools and how to assess information produced by them. The balance is especially relevant at DWP because the department administers benefits, pensions and employment support while handling substantial volumes of sensitive personal information.

Training does not determine how AI is deployed inside those services. Systems still require appropriate data controls, procurement, assurance, human oversight and routes for handling error, while completion of an introductory course does not establish competence to automate a complex administrative or decision-making process.

The £112,995 expenditure figure also needs a narrow interpretation. It covers the learning costs reported through the FOI rather than the full economic cost of building AI capability inside the department. Employee time, programme administration and project-specific technical development may sit elsewhere.

Government’s wider national target similarly measures the spread of skills rather than a specific productivity outcome. Large volumes of introductory training can create the conditions for adoption, but they do not show whether workers subsequently use approved tools more effectively or whether services improve.

DWP’s figures are useful precisely because they expose the shape of the programme. Tens of thousands of foundation-module completions demonstrate substantial exposure to basic AI concepts, while the much smaller volume of more specialised learning shows where broad literacy gives way to deeper capability.

The harder measurement comes next. Departments will need evidence showing whether trained staff adopt tools appropriately, make fewer mistakes, identify risks earlier or produce measurable service improvements. Until that evidence exists, 128,779 is best understood as a count of course completions rather than a proxy for the number of AI-skilled civil servants.