E26 UK AI Employment and Youth Model

National displacement missing job creation and international feedback

Desmond Last  |  Economics 2026  |  23 October 2026

economics2026.com

Purpose and main result

This report develops a five year quadratic scenario for the effect of artificial intelligence on UK employment. It brings together existing jobs removed, new jobs never created, reduced household spending, international feedback and a separate allocation to people aged 18 to 25.

Under the working assumptions, the model gives approximately 405,700 fewer people in UK employment by the end of 2031 relative to the counterfactual, including approximately 95,500 aged 18 to 25. These are conditional scenario results, not calibrated forecasts.

The clear working quadratics

National employment gap with provisional 10 per cent international feedback

Youth employment gap under the working age allocation

Time t is measured in years from the end of 2026. Thus t = 1 is the end of 2027 and t = 5 is the end of 2031. A positive gap means employment is lower than it would have been without the modelled AI effects.

How to read the results

Year end figures are employment shortfalls at that date. They must not be added across years as if they were different people. The youth figures are included within the national total. The model applies only over 2027 to 2031 and should not be extrapolated indefinitely.

An employment gap is not automatically the same as additional unemployment. Some people may instead remain in education, leave the labour force or change their working hours. An unemployment forecast needs those transitions and a separate economic baseline.

Domestic model and assumptions

The starting idea was to use the reported Barclays programme of around 5,000 job reductions as a benchmark for a national scenario. The announcement concerns a global bank and wider technology enabled restructuring. It does not establish an economy wide AI displacement rate or identify 5,000 UK AI redundancies. [1]

Input	Working value
UK employment starting scale N	34.577 million
Share encountering comparable restructuring q	50%
Programme benchmark z	5,000 / 93,000 = 5.376%
AI attribution within programme α	50%
Share remaining after replacement work r	50%
Domestic spending knock-on k	20%
Missing new roles over five years H(5)	150,000
Share without alternative employment h	50%

The employment starting scale and bank denominator are inherited inputs from the initial run, rather than newly established current measurements. The exposure, AI attribution, replacement and hiring assumptions are scenario choices. The bank ratio cannot identify both the speed and curvature of national adoption.

Define the assumed rollout path f(t), which reaches 1 in year five:

D(t) counts direct roles removed before replacement employment. New roles never created are represented separately:

This gives annual increments of 10,000, 20,000, 30,000, 40,000 and 50,000 missing roles. The reference pool of 100,000 potential roles a year was an illustrative hiring assumption, not a measured UK job creation total.

The displayed domestic polynomial is the numerical anchor for this report. Its coefficients were rounded in the initial run. Back calculation gives rD(t) = 23,237.2333t + 4,647.45t²; hH(t) = 2,500t + 2,500t². Recomputing from all original unrounded inputs can cause immaterial sub-person differences.

International feedback and government response

UK exports totalled approximately £910 billion and imports £929 billion in 2025, excluding precious metals. These totals establish the scale of international exposure; they do not establish the model’s international multiplier. [2]

X(t) is the UK employment gap from weaker foreign demand for UK exports. S(t) covers additional UK supplier losses caused by international supply disruption. V(t) covers reduced foreign investment, financial activity and corporate spending.

R(t) is further UK household spending feedback from those internationally generated losses. M(t) is the offset from lower import costs, productivity gains, inward investment and new foreign demand. All five terms must use the same employment units and counterfactual.

Allocate each effect once. Supplier effects already included in export or investment multipliers must not be added again under S(t). Household feedback already counted in R(t) must not receive a second spending multiplier. International losses overseas are counted only to the extent that they affect UK employment.

General quadratic form

G(t) is the net additional employment gap from government responses beyond the baseline, such as spending cuts, tax changes or employment support. Fiscal deterioration alone does not create this additional employment effect. The working run sets G(t) = 0.

Provisional international setting

The 10 per cent setting is a sensitivity assumption for net international feedback, including its household effects and offsets. It is not an evidence based UK multiplier. It forces the same timing as domestic effects; later calibration should estimate the two international coefficients independently.

Earlier illustrative international additions of 25 per cent and 50 per cent are retained as sensitivities, rather than alternative measured estimates. The international term can also be negative if beneficial effects exceed losses.

Estimating the domestic spending coefficient

Here s is the annual spending reduction per affected person, d is the share reaching UK production, and e is the UK employment effect per pound over the same period, including later spending rounds. This gives additional jobs affected per initial employment loss. Use the reduction in spending, not the entire former wage: benefits, savings and replacement earnings can sustain consumption.

National results and public finances

Year	Domestic gap	International gap	Total UK gap
2027	39,462	3,946	43,408
2028	96,077	9,608	105,685
2029	169,847	16,985	186,831
2030	260,770	26,077	286,847
2031	368,847	36,885	405,732

Figures are calculated before rounding each displayed result. The corrected 2030 national total is 286,847. Separately rounded columns can differ from their displayed sum by one person.

National composition at the end of 2031

Component	Employment gap
Displacement after replacement employment	232,372
Missing opportunities after alternative employment	75,000
Domestic spending knock-on	61,474
International feedback	36,885
Total from unrounded calculation	405,732

Annual financial effects at the 2031 level

The original financial assumptions are £10,000 less household spending, £8,000 less employment related tax and £6,000 more benefits per person in the gap per year. Applying them uniformly gives:

Effect	Annual amount
Household spending foregone	£4.06bn
Employment related tax lost	£3.25bn
Additional benefits	£2.43bn
Partial fiscal deterioration	£5.68bn

Fiscal deterioration is tax lost plus additional benefits. Household spending foregone is a separate measure and must not be added to the fiscal total. These are annual amounts at the 2031 gap, not cumulative five year costs.

These averages are assumptions, not tax or benefit entitlement calculations. A person in a missing job may receive family support or remain in education. Corporation tax, VAT effects, investment changes, borrowing costs and fiscal effects of AI gains require separate accounts. The spending estimate describes the modelled effect; it is not another shock to add to the jobs calculation.

Effects on people aged 18 to 25

Young people may experience both existing job displacement and fewer opportunities to enter work. The Bank of England highlights uncertainty around graduate and junior roles. Government analysis of LinkedIn data finds hiring weakness and a preference for experienced candidates in declining entry level occupations, while cautioning against attributing the decline conclusively to AI. [3, 4]

The working youth allocation assumes that ages 18 to 25 bear 15 per cent of the displacement remaining after replacement employment and 50 per cent of missing opportunities remaining after alternative employment. Both shares are testing assumptions. Entry level work is not exclusive to this age group.

This allocation also assumes a 20 per cent household knock-on and 10 per cent international addition to the youth base. These must eventually be replaced by age specific transmission estimates. Missing new opportunities are mutually exclusive from displaced workers in this accounting.

Year	Youth employment gap
2027	8,821
2028	22,783
2029	41,885
2030	66,127
2031	95,510

Youth composition at the end of 2031

Component	Youth gap
Existing employment lost after replacement work	34,856
Missing new opportunities after alternative employment	37,500
Domestic household spending knock-on	14,471
International feedback	8,683
Total	95,510

Before rounding, the first two components total 72,355.8625. Multiplying by 1.20 and then 1.10 gives 95,509.7385, or approximately 95,510. The youth gap is 23.5 per cent of the national modelled gap. This is not the youth share of all unemployment or the youth unemployment rate.

Each year refers to whoever is aged 18 to 25 at that date. The model does not follow a fixed cohort as it ages. Keeping the allocation consistent over time requires explicit age entry and exit flows in the next version.

Sensitivity and implications for young people

The sensitivity range changes the allocation of the national displacement and missing opportunity components to ages 18 to 25. It retains the same domestic and international feedback settings.

Youth scenario	Displacement share	Missing opportunity share	2031 gap
Lower allocation	10%	30%	60,373
Working allocation	15%	50%	95,510
Higher allocation	20%	70%	130,646

This is an illustrative sensitivity range of approximately 60,000 to 131,000, not a confidence interval or an evidence based range of likely outcomes.

National sensitivity to international feedback

Net international addition	2031 national gap
0%	368,847
10% working setting	405,732
25%	461,059
50%	553,270

The route into employment

The E26 concern is that automation could remove the first rung of the employment ladder. Reduced junior recruitment can delay the acquisition of experience, intensify competition for remaining positions and increase dependence on benefits or family support. Reduced apprenticeships are a risk to test, rather than an established output of this model.

Delayed entry into work could also lower early earnings, spending and tax contributions. Persistent earnings damage as young people age requires a separate cohort model. It is not quantified by the current quadratic and should not be added as another headcount loss.

AI may also create roles, support learning and improve productivity. The replacement factors and net international offset allow for some benefits, but do not comprehensively quantify domestic investment, lower prices or new industries. A full net impact model needs those channels explicitly.

Working E26 statement

Under the provisional assumptions, the working scenario produces approximately 405,700 fewer people in UK employment by the end of 2031 relative to the counterfactual. Around 95,500 of that gap is allocated to ages 18 to 25. The purpose is to test the consequences of displacement and lost entry opportunities while identifying what evidence is needed to replace assumptions.

Calibration and sources

The next version should estimate national and age specific coefficients from sector and trading partner data. It should test the acceleration of adoption rather than assume a quadratic beyond its five year scope.

Evidence needed for calibration

Measure adoption, occupational tasks, redundancies, normal turnover, replacement hiring and genuinely new positions by sector and age. Separate gross hiring from net job creation, and distinguish an AI effect from weaker demand or other restructuring.

Build a sector by country matrix of foreign demand shocks, UK output exposure and employment effects. A demand channel can be calculated as the sum across sectors of demand losses in pounds multiplied by the appropriate UK employment effect per pound. Use matching price years and demand definitions.

ONS publishes direct FTE employment per £million and Type 1 employment effects and multipliers. These are useful foundations, but represent FTE jobs rather than unemployed people. Type 1 effects include inter-industry supply linkages; avoid multiplying in a second UK production share or adding supplier losses already covered. Household induced effects need separate treatment. [5]

Estimate supply disruption, foreign investment and beneficial offsets independently. Convert FTE effects into people using sector and age specific working hours, re-employment and labour-force transitions. Government response and permanent cohort effects should be separate modules.

Sources and provenance

[1] Reuters, 22 October 2026. Barclays to slash 5,000 jobs in AI-powered cost-cutting drive. Reports the bank programme; it does not calibrate the national or youth coefficients.
https://www.reuters.com/business/finance/barclays-q3-profit-rises-12-announces-670-million-buyback-2026-10-22/

[2] House of Commons Library, 16 October 2026. Trade in goods and services Economic indicators. 2025 trade totals exclude precious metals.
https://commonslibrary.parliament.uk/research-briefings/sn02815/

[3] Bank of England. How will AI affect jobs and wages. Discusses entry level work and uncertainty over causation and scale.
https://www.bankofengland.co.uk/explainers/how-will-ai-affect-jobs-and-wages

[4] Department for Science Innovation and Technology, 8 June 2026. Entry-level hiring in the UK A snapshot. LinkedIn hiring data are not an estimate of AI caused youth job losses.
https://www.gov.uk/government/publications/entry-level-hiring-in-the-uk-a-snapshot

[5] Office for National Statistics, 19 December 2025. Employment multipliers and effects in the UK. Includes a 2023 edition; official statistics in development.
https://www.ons.gov.uk/economy/nationalaccounts/supplyandusetables/datasets/employmentmultipliersandeffectsintheuk2020

Model provenance: E26 scenario developed by Desmond Last with AI assistance. Numerical assumptions are identified in the report; external sources provide context and calibration methods rather than validating the scenario outputs.