UK AI Jobs Tracker
34.9% of UK businesses with 10 or more employees used at least one AI technology in June 2026, up from 11.9% in September 2023, according to the Office for National Statistics. On Indeed, 9.66% of UK job postings mentioned AI on 31 August 2026, against 4.90% a year earlier, while total postings were lower than a year before. None of these sources measures job losses: they track business adoption, advertised demand, how people use AI tools and where AI capability overlaps with work.
Data last updated: 15 September 2026 Sources checked weekly
UK headline figures
Four measures, each from a different source and each describing a different population. They are not combined into a single score.
UK businesses with 10 or more employees using at least one AI technology
34.9%
Share of businesses, weighted by count. Wave 159.
ONS Business Insights and Conditions Survey, 15 to 28 June 2026
UK job postings on Indeed that mention AI
9.66%
Up from 4.90% on 31 August 2025, a rise of 4.76 percentage points.
Indeed Hiring Lab AI Tracker, seven-day average to 31 August 2026
Claude.ai use per working-age person in the UK, against the global average
3.35x
An index where the global average is 1. It covers Claude.ai users only.
Anthropic Economic Index, May 2026
UK consumer ChatGPT messages that are work-related
32.3%
Share of sampled messages. Consumer ChatGPT only: excludes enterprise accounts and Codex.
OpenAI Signals, June 2026
How many UK businesses use AI
The Office for National Statistics asks UK businesses about AI in its Business Insights and Conditions Survey about once a quarter. It counts light and embedded use the same as extensive use, so the chart measures breadth of adoption rather than depth.
UK businesses with 10 or more employees using at least one AI technology
Base (ONS wording): As a percentage of businesses not permanently stopped trading, broken down by industry and size band, weighted by count, UK.
Calculated by The AI Consultancy as all businesses minus those answering “does not currently use” or “not sure”, the method of the ONS article.
View the data
| Survey wave | Fieldwork | Share |
|---|---|---|
| Wave 92 | 18 September to 1 October 2023 | 11.9% |
| Wave 98 | 11 to 24 December 2023 | 11.8% |
| Wave 105 | 18 to 31 March 2024 | 13.8% |
| Wave 111 | 17 to 30 June 2024 | 14.8% |
| Wave 117 | 16 to 29 September 2024 | 17.8% |
| Wave 123 | 16 to 29 December 2024 | 18.2% |
| Wave 129 | 17 to 30 March 2025 | 20.6% |
| Wave 135 | 16 to 29 June 2025 | 25.1% |
| Wave 141 | 15 to 28 September 2025 | 27.2% |
| Wave 147 | 15 to 28 December 2025 | 28.7% |
| Wave 153 | 16 to 29 March 2026 | 32.1% |
| Wave 159 | 15 to 28 June 2026 | 34.9% |
Adoption figures describe the market as a whole. To estimate what AI could save in your own operations, use the AI ROI calculator.
ONS also asks what AI has done, or is expected to do, to headcount. Each question has its own base, printed with its panel: most are asked only of businesses already using AI.
What UK businesses with 10 or more employees report about AI and headcount
Effect of AI on headcount so far
Base (ONS wording): As a percentage of businesses not permanently stopped trading that are using some form of artificial intelligence or are not sure if they are doing so, broken down by industry and size band, weighted by count, UK.
Expected effect on headcount of planned AI use
Base (ONS wording): As a percentage of businesses not permanently stopped trading that plan to use artificial intelligence within the next three months or are not sure about doing so, broken down by industry and size band, weighted by count, UK.
Effect of AI on headcount so far
Base (ONS wording): As a percentage of businesses not permanently stopped trading that are using some form of artificial intelligence or are not sure if they are doing so, broken down by industry and size band, weighted by count, UK.
Expected effect on headcount of planned AI use
Base (ONS wording): As a percentage of businesses not permanently stopped trading that plan to use artificial intelligence within the next three months or are not sure about doing so, broken down by industry and size band, weighted by count, UK.
Not published by ONS for this size band in wave 159: “Increased headcount”.
These answers are businesses’ own reports of headcount effects.
View the data
| Question and answer | Share |
|---|---|
| Effect of AI on headcount so far: No change | 56.6% |
| Effect of AI on headcount so far: Reduced headcount | 5.0% |
| Effect of AI on headcount so far: Not sure | 20.4% |
| Effect of AI on headcount so far: Not applicable | 17.1% |
| Expected effect on headcount of planned AI use: Expect headcount to increase | 1.2% |
| Expected effect on headcount of planned AI use: Expect no change | 51.0% |
| Expected effect on headcount of planned AI use: Expect headcount to reduce | 9.9% |
| Expected effect on headcount of planned AI use: Not sure | 37.8% |
What UK employers are hiring for
Indeed Hiring Lab tracks the share of job postings that mention AI-related terms in 9 countries. On 31 August 2026 the UK ranked third of the 9, behind Canada and Ireland.
Total job postings on Indeed in the UK
The AI share above is a share of these postings, so it can rise when total postings fall. The index stood at 68.41 on 11 September 2026, against 74.78 a year earlier.
View the data
| Date | Index |
|---|---|
| 29 February 2020 | 99.48 |
| 31 March 2020 | 61.01 |
| 30 April 2020 | 38.77 |
| 31 May 2020 | 34.37 |
| 30 June 2020 | 36.82 |
| 31 July 2020 | 43.09 |
| 31 August 2020 | 49.19 |
| 30 September 2020 | 52.18 |
| 31 October 2020 | 57.48 |
| 30 November 2020 | 62.43 |
| 31 December 2020 | 68.56 |
| 31 January 2021 | 62.49 |
| 28 February 2021 | 66.61 |
| 31 March 2021 | 84.82 |
| 30 April 2021 | 99.27 |
| 31 May 2021 | 116.63 |
| 30 June 2021 | 123.11 |
| 31 July 2021 | 134.49 |
| 31 August 2021 | 143.65 |
| 30 September 2021 | 150.32 |
| 31 October 2021 | 158.49 |
| 30 November 2021 | 161.74 |
| 31 December 2021 | 153.20 |
| 31 January 2022 | 157.49 |
| 28 February 2022 | 165.22 |
| 31 March 2022 | 170.89 |
| 30 April 2022 | 164.08 |
| 31 May 2022 | 171.23 |
| 30 June 2022 | 164.69 |
| 31 July 2022 | 164.25 |
| 31 August 2022 | 164.52 |
| 30 September 2022 | 157.22 |
| 31 October 2022 | 159.10 |
| 30 November 2022 | 156.66 |
| 31 December 2022 | 149.60 |
| 31 January 2023 | 145.85 |
| 28 February 2023 | 139.40 |
| 31 March 2023 | 137.40 |
| 30 April 2023 | 134.03 |
| 31 May 2023 | 130.60 |
| 30 June 2023 | 127.33 |
| 31 July 2023 | 125.24 |
| 31 August 2023 | 123.98 |
| 30 September 2023 | 121.64 |
| 31 October 2023 | 117.79 |
| 30 November 2023 | 112.66 |
| 31 December 2023 | 108.90 |
| 31 January 2024 | 105.94 |
| 29 February 2024 | 104.83 |
| 31 March 2024 | 104.56 |
| 30 April 2024 | 101.30 |
| 31 May 2024 | 99.75 |
| 30 June 2024 | 96.07 |
| 31 July 2024 | 92.88 |
| 31 August 2024 | 89.76 |
| 30 September 2024 | 88.41 |
| 31 October 2024 | 86.10 |
| 30 November 2024 | 86.50 |
| 31 December 2024 | 88.06 |
| 31 January 2025 | 84.85 |
| 28 February 2025 | 83.09 |
| 31 March 2025 | 81.82 |
| 30 April 2025 | 75.34 |
| 31 May 2025 | 78.32 |
| 30 June 2025 | 76.63 |
| 31 July 2025 | 77.13 |
| 31 August 2025 | 74.59 |
| 30 September 2025 | 75.74 |
| 31 October 2025 | 75.66 |
| 30 November 2025 | 76.98 |
| 31 December 2025 | 76.82 |
| 31 January 2026 | 75.41 |
| 28 February 2026 | 76.85 |
| 31 March 2026 | 72.86 |
| 30 April 2026 | 69.91 |
| 31 May 2026 | 67.11 |
| 30 June 2026 | 66.46 |
| 31 July 2026 | 68.47 |
| 31 August 2026 | 69.07 |
| 11 September 2026 | 68.41 |
How people use AI at work
OpenAI and Anthropic publish data on how people use their own products, so these figures cover ChatGPT and Claude users only.
Share of consumer ChatGPT messages that are work-related, UK and global
- UK
- Global
A falling work share means non-work messages grew faster than work messages; on its own it does not show work use shrinking.
View the data
| Month | UK | Global |
|---|---|---|
| July 2024 | 62.3% | 51.3% |
| August 2024 | 59.0% | 48.3% |
| September 2024 | 55.7% | 46.6% |
| October 2024 | 55.2% | 45.4% |
| November 2024 | 55.0% | 43.7% |
| December 2024 | 49.1% | 40.1% |
| January 2025 | 50.9% | 40.3% |
| February 2025 | 47.3% | 38.9% |
| March 2025 | 48.0% | 37.6% |
| April 2025 | 44.1% | 35.2% |
| May 2025 | 41.5% | 35.5% |
| June 2025 | 38.4% | 33.3% |
| July 2025 | 36.0% | 31.4% |
| August 2025 | 34.1% | 30.6% |
| September 2025 | 36.2% | 33.0% |
| October 2025 | 38.0% | 33.9% |
| November 2025 | 37.4% | 33.6% |
| December 2025 | 33.4% | 30.5% |
| January 2026 | 36.3% | 30.7% |
| February 2026 | 34.6% | 30.8% |
| March 2026 | 35.0% | 31.0% |
| April 2026 | 34.3% | 31.7% |
| May 2026 | 32.6% | 30.5% |
| June 2026 | 32.3% | 30.4% |
In May 2026, 55.2% of UK Claude.ai conversations followed an augmentation pattern (iterating, learning or checking work with Claude) and 44.8% an automation pattern (directing Claude to complete a task).
UK Claude.ai use by occupation group, split by interaction pattern
- Automation pattern
- Augmentation pattern
Patterns describe how people interact with Claude, not whether jobs are automated. Conversations are assigned to groups by the tasks they involve. These 10 groups account for 89.5% of UK Claude.ai use.
View the data
| Occupation group (US SOC) | Share of use | Automation pattern | Augmentation pattern |
|---|---|---|---|
| Computer and Mathematical | 20.21% | 63.09% | 36.91% |
| Educational Instruction and Library | 15.26% | 35.53% | 64.47% |
| Arts, Design, Entertainment, Sports, and Media | 13.53% | 29.15% | 70.85% |
| Sales and Related | 9.94% | 43.66% | 56.34% |
| Office and Administrative Support | 7.27% | 55.51% | 44.49% |
| Management | 5.94% | 41.26% | 58.74% |
| Business and Financial Operations | 5.80% | 39.51% | 60.49% |
| Life, Physical, and Social Science | 4.44% | 36.75% | 63.25% |
| Healthcare Practitioners and Technical | 3.94% | 42.18% | 57.82% |
| Community and Social Service | 3.20% | 37.67% | 62.33% |
Which kinds of work AI could already do
The OECD AI Exposure Measure compares what each occupation demands, in domains such as language, problem solving and social interaction, with what current AI can do. A high score means AI capability overlaps with what the work demands.
Potential AI exposure by US occupation group
This is potential exposure, not job loss. Groups whose work makes low demands in some domains can score highly even where the work is manual. Occupations use US definitions.
View the data
| Occupation group (US SOC 2018) | SOC code | Mean index |
|---|---|---|
| Office and Administrative Support | 43 | 0.968 |
| Production | 51 | 0.924 |
| Food Preparation and Serving Related | 35 | 0.896 |
| Sales and Related | 41 | 0.888 |
| Building and Grounds Cleaning and Maintenance | 37 | 0.883 |
| Farming, Fishing, and Forestry | 45 | 0.882 |
| Transportation and Material Moving | 53 | 0.880 |
| Healthcare Support | 31 | 0.878 |
| Personal Care and Service | 39 | 0.870 |
| Construction and Extraction | 47 | 0.865 |
| Computer and Mathematical | 15 | 0.864 |
| Architecture and Engineering | 17 | 0.853 |
| Installation, Maintenance, and Repair | 49 | 0.847 |
| Business and Financial Operations | 13 | 0.840 |
| Arts, Design, Entertainment, Sports, and Media | 27 | 0.823 |
| Life, Physical, and Social Science | 19 | 0.791 |
| Protective Service | 33 | 0.774 |
| Management | 11 | 0.762 |
| Healthcare Practitioners and Technical | 29 | 0.755 |
| Educational Instruction and Library | 25 | 0.748 |
| Legal | 23 | 0.746 |
| Community and Social Service | 21 | 0.723 |
Sector read
How the evidence lines up for 11 sectors The AI Consultancy works with, using US occupations mapped to each sector. Evidence strength is our rating of how much published data covers them; none is rated strong, because every source uses US occupational definitions.
| Sector | Evidence | OECD exposure | Observed exposure | Share of Claude.ai use | Automation pattern | Our read |
|---|---|---|---|---|---|---|
| Recruitment and professional staffing | Limited | 0.839 | 0.234 | 0.08% | 42.1% | Evidence is thin. Base decisions on a review of your own workflows, not sector figures. |
| Legal and professional services | Moderate | 0.871 | 0.167 | 1.55% | 38.8% | Augmentation first: research, drafting and review, with accountable human sign-off. |
| Finance and investment | Moderate | 0.854 | 0.234 | 1.83% | 45.5% | Separate transaction processing, where automation fits, from advice and compliance, which need sign-off. |
| Property and real estate | Limited | 0.829 | 0.213 | 0.12% | 49.5% | Too little task-level usage data to support sector claims. |
| Customer service | Moderate | 0.960 | 0.386 | 1.03% | 41.8% | Exposure is high, but observed use leans towards AI-assisted responses with escalation, not unattended handling. |
| Administrative operations | Moderate | 0.995 | 0.213 | 3.62% | 62.4% | The clearest case for automation-oriented reviews of records, data entry and formatting work. |
| Marketing and content | Moderate | 0.830 | 0.283 | 11.43% | 33.6% | Heavy use that leans towards augmentation: capacity and quality gains, with editorial judgement kept human. |
| Software and IT | Moderate | 0.901 | 0.337 | 22.01% | 65.8% | The heaviest, most automation-leaning use; coding agents sit largely outside this data. |
| Education | Moderate | 0.730 | 0.099 | 12.85% | 42.6% | Heavy use alongside comparatively low OECD exposure: support tasks around teaching are the likelier fit. |
| Construction | Limited | 0.866 | 0.000 | 0.11% | 64.7% | Usage is close to zero; relevance lies in estimating, records and coordination, not site work. |
| Logistics | Limited | 0.935 | 0.008 | 0.56% | 48.1% | OECD exposure is high partly because manual demands are low; opportunities sit in clerical and planning tasks. |
Recruitment and professional staffing
- Evidence
- Limited
- OECD exposure (median)
- 0.839
- Observed exposure (median)
- 0.234
- Share of global Claude.ai use
- 0.08%
- Automation pattern
- 42.1%
Evidence is thin. Base decisions on a review of your own workflows, not sector figures.
Legal and professional services
- Evidence
- Moderate
- OECD exposure (median)
- 0.871
- Observed exposure (median)
- 0.167
- Share of global Claude.ai use
- 1.55%
- Automation pattern
- 38.8%
Augmentation first: research, drafting and review, with accountable human sign-off.
Related: Legal, Professional services
Finance and investment
- Evidence
- Moderate
- OECD exposure (median)
- 0.854
- Observed exposure (median)
- 0.234
- Share of global Claude.ai use
- 1.83%
- Automation pattern
- 45.5%
Separate transaction processing, where automation fits, from advice and compliance, which need sign-off.
Related: Financial services
Property and real estate
- Evidence
- Limited
- OECD exposure (median)
- 0.829
- Observed exposure (median)
- 0.213
- Share of global Claude.ai use
- 0.12%
- Automation pattern
- 49.5%
Too little task-level usage data to support sector claims.
Customer service
- Evidence
- Moderate
- OECD exposure (median)
- 0.960
- Observed exposure (median)
- 0.386
- Share of global Claude.ai use
- 1.03%
- Automation pattern
- 41.8%
Exposure is high, but observed use leans towards AI-assisted responses with escalation, not unattended handling.
Administrative operations
- Evidence
- Moderate
- OECD exposure (median)
- 0.995
- Observed exposure (median)
- 0.213
- Share of global Claude.ai use
- 3.62%
- Automation pattern
- 62.4%
The clearest case for automation-oriented reviews of records, data entry and formatting work.
Marketing and content
- Evidence
- Moderate
- OECD exposure (median)
- 0.830
- Observed exposure (median)
- 0.283
- Share of global Claude.ai use
- 11.43%
- Automation pattern
- 33.6%
Heavy use that leans towards augmentation: capacity and quality gains, with editorial judgement kept human.
Software and IT
- Evidence
- Moderate
- OECD exposure (median)
- 0.901
- Observed exposure (median)
- 0.337
- Share of global Claude.ai use
- 22.01%
- Automation pattern
- 65.8%
The heaviest, most automation-leaning use; coding agents sit largely outside this data.
Education
- Evidence
- Moderate
- OECD exposure (median)
- 0.730
- Observed exposure (median)
- 0.099
- Share of global Claude.ai use
- 12.85%
- Automation pattern
- 42.6%
Heavy use alongside comparatively low OECD exposure: support tasks around teaching are the likelier fit.
Construction
- Evidence
- Limited
- OECD exposure (median)
- 0.866
- Observed exposure (median)
- 0.000
- Share of global Claude.ai use
- 0.11%
- Automation pattern
- 64.7%
Usage is close to zero; relevance lies in estimating, records and coordination, not site work.
Logistics
- Evidence
- Limited
- OECD exposure (median)
- 0.935
- Observed exposure (median)
- 0.008
- Share of global Claude.ai use
- 0.56%
- Automation pattern
- 48.1%
OECD exposure is high partly because manual demands are low; opportunities sit in clerical and planning tasks.
Related: Logistics and transport
Sources: OECD AI Exposure Measure; Anthropic Economic Index, including its labour market impacts data for observed exposure. Sector mapping and medians by The AI Consultancy.
What this data does and does not tell you
- Exposure is not automation. The OECD measure shows where AI capability overlaps with what work demands, not what employers have automated.
- Automation is not job loss. An automation-pattern conversation with Claude, or a business automating a task, does not mean a role has gone.
- Provider data describes that provider’s users. ChatGPT and Claude figures cover their users, not the UK labour market.
- Job adverts that mention AI are advertised demand, not adoption. A posting can mention AI without being an AI role.
- A share of businesses is not a share of workers. ONS weights by business count, so a small firm counts the same as a large one.
- Occupations use US definitions. Neither the OECD nor the Anthropic measure is published using UK occupational definitions.
Methodology and sources
Every figure on this page is generated from The AI Consultancy’s research repository, which keeps each source’s published files and checks weekly for new releases. No figure is typed by hand, and sources are never merged into one score.
Office for National Statistics: Business Insights and Conditions Survey (BICS): AI questions
- Measure
- Official survey estimate: share of UK businesses, weighted by business count
- Population
- UK businesses with 10 or more employees, weighted by count
- Release
- BICS wave 159 (reissued 7 July 2026), dated 7 July 2026. Data: 18 September 2023 to 28 June 2026.
- Changes made
- Share using at least one AI technology calculated by The AI Consultancy as 1 minus 'does not currently use' minus 'not sure'; shares converted to percentages.
- Limitations
- Proportions of businesses weighted by count: they describe businesses, not workers, employment or intensity of use. ONS headlines businesses with 10 or more employees because the weights for micro businesses (0 to 9 employees) can have a disproportionate influence on the all-business figure.
- AI use is measured as a list of technologies, so light and embedded use count the same as extensive use.
- Workforce and headcount questions are business-reported perceptions, not measured employment change.
- Attribution
- Source: Office for National Statistics, Business Insights and Conditions Survey. Contains public sector information licensed under the Open Government Licence v3.0. Licence
Indeed Hiring Lab: AI Tracker
- Measure
- Job postings platform data: share of job postings on Indeed mentioning AI, seven-day trailing average
- Population
- Job postings on Indeed in each country the tracker covers (all occupations)
- Release
- AI_posting.csv, data to 31 August 2026, dated 10 September 2026. Data: 1 January 2019 to 31 August 2026.
- Changes made
- Month-end values selected from the daily series; values rounded to two decimal places.
- Limitations
- Measures employer demand advertised on Indeed, not AI adoption, use of AI at work, productivity or job losses.
- Share of postings, not of vacancies or employment; a rising share can reflect falling total postings, so read with the postings index.
- Hiring Lab revises the full history at each update: values quoted from an earlier file (or a Hiring Lab article) can differ from the current file without any labour-market change.
- Attribution
- Indeed Hiring Lab, AI Tracker, https://github.com/hiring-lab/ai-tracker, licensed under CC BY 4.0. Licence
Indeed Hiring Lab: Job Postings Index (GB)
- Measure
- Job postings platform data: index of total job postings on Indeed, seasonally adjusted, 1 February 2020 = 100
- Population
- All job postings on Indeed in GB
- Release
- aggregate_job_postings_GB.csv, data to 11 September 2026, dated 15 September 2026. Data: 1 February 2020 to 11 September 2026.
- Changes made
- Month-end values selected from the daily series.
- Limitations
- Hiring Lab revises the full history at each update: values quoted from an earlier file (or a Hiring Lab article) can differ from the current file without any labour-market change.
- Deduplication is designed for the job seeker experience, not accurate counting (Hiring Lab data FAQ).
- Attribution
- Indeed Hiring Lab, Job Postings Index, https://github.com/hiring-lab/job_postings_tracker, licensed under CC BY 4.0. Licence
OpenAI: OpenAI Signals v2.0
- Measure
- Observed usage: share of sampled consumer ChatGPT messages classified as work-related
- Population
- Consumer ChatGPT messages (Free, Plus, Pro and Go), adult accounts; excludes enterprise and Codex
- Release
- OpenAI Signals v2.0, dated 6 August 2026. Data: July 2024 to June 2026.
- Changes made
- Shares converted to percentages.
- Limitations
- Consumer ChatGPT only: understates business use (stated by OpenAI).
- Differentially private shares rounded to 3 decimals (5 for O*NET IWA); small cells are noisy.
- v2.0 restated earlier periods: historical values can change between releases without any labour-market change.
- Attribution
- Chatterji, Cunningham, Deming, Hitzig, Johnston, Richmond, Ong, Shan and Wadman, "OpenAI Signals v2.0", https://cdn.openai.com/signals/data-dictionary.pdf, licensed under CC BY 4.0. Licence
Anthropic: Anthropic Economic Index (release of 26 June 2026)
- Measure
- Observed usage: share of Claude.ai conversations, and the interaction pattern (automation or augmentation) of that use
- Population
- Claude.ai chat and Cowork conversations (Free, Pro and Max plans)
- Release
- release_2026_06_26 (6th report, 'Cadences'), dated 26 June 2026. Data: 1 to 31 May 2026.
- Changes made
- Top ten UK occupation groups selected by share of use; sector figures aggregated by The AI Consultancy.
- Limitations
- Claude users only; skewed towards computer and mathematical work; not representative of the labour market.
- Occupation assignment is inferred from conversation content mapped to O*NET tasks, not from users' actual jobs.
- Automation/augmentation describes how people interact with Claude, not whether jobs are automated.
- Attribution
- Massenkoff, Lyubich, Sacher, Hitzig, Zhang, Heller and McCrory, "Anthropic Economic Index report: Cadences", 26 June 2026. Data released under CC BY. Licence
Anthropic: Anthropic Economic Index: labour market impacts data (job exposure)
- Measure
- Usage-weighted exposure: observed exposure score per US occupation, 0 to 1
- Population
- US SOC occupations; usage from Anthropic Economic Index samples (August and November 2025)
- Release
- labour market impacts data, dated 5 March 2026. Data: Single release, not a time series.
- Changes made
- Sector medians calculated by The AI Consultancy.
- Limitations
- Claude users only; skewed towards computer and mathematical work; not representative of the labour market.
- Observed exposure weights automated use at full value and augmented use at half value (Anthropic judgement calls, stated in the report).
- Attribution
- Anthropic, Anthropic Economic Index, labour market impacts data (job_exposure.csv), 5 March 2026, https://huggingface.co/datasets/Anthropic/EconomicIndex. Data released under CC BY. Licence
OECD: The OECD AI Exposure Measure (AI Capability Gap Index)
- Measure
- Capability exposure: reversed normalised AI capability gap index, higher means more exposed
- Population
- US O*NET-SOC occupations with Skills, Abilities and Work Context data; no employment weighting
- Release
- OECD Artificial Intelligence Papers No. 59, dated 26 May 2026. Data: Single release, not a time series.
- Changes made
- Occupation values averaged by US SOC major group (unweighted mean) and summarised by sector (median) by The AI Consultancy.
- Limitations
- Measures potential exposure (capability versus occupational requirements), not usage, adoption or employment effects.
- Reversed normalised index is provided 'for comparison reasons'; the total gap index is the primary measure (lower = more exposed).
- The total index weights each domain gap by its importance to the occupation; OECD states rankings should be read together with domain-level results, and low-demand manual occupations can score as highly exposed.
- Attribution
- OECD (2026), The OECD AI Exposure Measure: Mapping the OECD AI Capability Indicators to Occupations, OECD Artificial Intelligence Papers, No. 59, OECD Publishing, Paris, https://doi.org/10.1787/f3da0f0a-en, licensed under CC BY 4.0. Licence
OECD adaptation notice
This is an adaptation of an original work by the OECD. The opinions expressed and arguments employed in this adaptation should not be reported as representing the official views of the OECD or of its Member countries.
O*NET credit
This page includes information from the O*NET 31.0 Database by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA). Used under the CC BY 4.0 license. O*NET® is a trademark of USDOL/ETA. The AI Consultancy has modified all or some of this information. USDOL/ETA has not approved, endorsed, or tested these modifications. O*NET database | CC BY 4.0
Update log
Data last updated 15 September 2026, the date of the newest source release included. Sources are checked weekly; the page changes only when a source publishes new data.
: Tracker launched
First publication, using the latest releases from the Office for National Statistics, Indeed Hiring Lab, OpenAI Signals, the Anthropic Economic Index and the OECD AI Exposure Measure.
Frequently asked questions
Which jobs are most at risk from AI in the UK?
None of the sources in this tracker measures which UK jobs are at risk, so it reports exposure. On the OECD AI Exposure Measure, Office and Administrative Support is the most exposed of 22 US occupation groups (0.968) and Community and Social Service the least (0.723). Exposure means AI capability overlaps with the work; it is not a forecast of automation or job loss.
Is AI reducing jobs in the UK?
The data does not show that directly. In ONS survey wave 159, among UK businesses with 10 or more employees using AI or unsure whether they do, 5.0% reported reduced headcount, 56.6% no change and 20.4% were not sure. These are businesses' own reports, not measured employment change.
How many UK businesses use AI?
34.9% of UK businesses with 10 or more employees used at least one AI technology in ONS survey wave 159 (15 to 28 June 2026), up from 11.9% in wave 92 (18 September to 1 October 2023). It is a share of businesses, not of workers.
Are more UK job adverts mentioning AI?
Yes. 9.66% of UK job postings on Indeed mentioned AI on 31 August 2026, against 4.90% a year earlier and 6.74% in the US. A posting counts when it mentions AI-related keywords, so this is advertised demand, not a count of AI roles.
How often is the UK AI Jobs Tracker updated?
When a source publishes new data. Sources are checked weekly; the newest release included is dated 15 September 2026.
The tracker changes when a source publishes new data. Subscribe to get the update when new data lands.
Get new briefings by email
The AI Consultancy newsletter delivers briefings and analysis for UK businesses. We use your address only to send it, and you can unsubscribe at any time.