UK AI Displacement & Levy Model

If AI takes the jobs,
who pays for what comes next?

Use the sliders to explore what happens when AI automates UK work. The charts update live.

Simple view — 3-step walkthrough
Step 1 — The Problem

AI can automate 35% of UK work tasks. At current rates, that puts 5.4M jobs at risk of significant disruption.

How much work can AI automate? 35%
Low (5%)High (65%)
Step 2 — The Cost

When those workers stop paying tax, the government loses £2.9B/year in income tax and National Insurance — a hole that opens faster than new jobs appear.

Calculated automatically from Step 1 · avg wage £38,000 · 28% effective rate · ONS 2025

Step 3 — The Bridge

A 5% levy on commercial AI usage would raise £0.5B/year — covering 17% of that gap.

What % levy on commercial AI usage? 5%
0%15%
Your scenario: —
How does this play out over time?
Jobs displaced vs reinstated vs levy revenue (2025–2032). The gap in the middle is what we’re trying to bridge.
Displaced jobs (M)
Reinstated jobs (M)
Levy revenue (£B)
Fiscal gap (£B)

Model Parameters

Displacement Scenario
How much of work can AI actually do?
Automation depth35%
Of that, how much leads to actual job losses?
Displacement rate45%
What % levy on commercial AI usage?
Levy rate5.0%
How fast will new jobs appear to replace the ones lost?
Reinstatement Rate (ρ)40%
How much will AI boost the overall economy?
Productivity Multiplier (λ)1.30
How quickly are UK businesses actually adopting AI?
Adoption SpeedModerate
Token Tax Structure
Who pays — AI companies, or the businesses using AI?
Tax Incidence Split50/50
At what usage level does the tax kick in?
SME Exemption Threshold1M
What should the money raised be spent on?
Revenue RecyclingRetraining
UK Baseline Figures (ONS 2025)
Employed workforce: 34.24M
Employment rate (16–64): 75.0%
Avg nominal wage growth: 4.2%
Income tax + NI receipts: ~£440B/yr
UK AI VC raised (2025): £6B
New to this? Use the red sliders on the left to set how fast you think AI will displace work. Then try the blue Levy Rate slider. The model's central question: can the purple number (annual levy revenue) cover the red number (peak fiscal gap)?
Jobs Displaced (UK)
—
workers whose tasks get automated
Net Jobs (after reinstatement)
—
new roles created, minus roles lost
UK GDP Impact
—
net economy-wide effect at full adoption
Trough Fiscal Gap
—
annual tax revenue lost at peak disruption
Token Levy Revenue
—
raised annually by taxing AI usage
Fiscal Gap Covered
—
how much the levy closes the gap
Producer Burden
—
paid by AI companies (OpenAI etc.)
Consumer Burden
—
paid by businesses that buy AI services
Transition Trough — Employment & Fiscal Timeline
Parametric model: displacement curve vs reinstatement curve vs levy revenue. Based on Acemoglu–Restrepo displacement/reinstatement decomposition calibrated to IPPR UK scenarios.
Cumulative displaced jobs (M)
Reinstated jobs (M)
Net employment change (M)
Levy revenue (£B/yr)
Fiscal gap (£B/yr)
IPPR Scenario Comparison
Your model vs IPPR published scenarios (2024). Jobs lost shown as bars; GDP gain as label.
Levy Rate Sensitivity
Annual revenue at different levy rates for current adoption level. Shaded = SME exemption impact.
GDP Composition — Displacement Effect vs Productivity Effect
Acemoglu–Restrepo decomposition: automation raises productivity but the displacement effect partially offsets it. Net GDP = productivity gain minus displacement drag minus transition cost, plus reinstatement uplift.
Model interpretation: The productivity effect (λ−1) raises GDP through cost savings and output expansion. The displacement effect (−α×δ×labour share) reduces it by lowering wage income and thus consumption. The reinstatement effect (ρ×displaced) partially restores it via new tasks. Net fiscal impact equals the change in the income tax + NI base. The token levy acts as a bridging mechanism during the trough.
ONS LFS Feb 2026 IPPR (2024) DSIT AI Sector Study 2024 GOV.UK AI Labour Impact Jan 2026 PwC UK AI Report Acemoglu–Restrepo (2019, 2022) Goldman Sachs (2025) ONS BICS Sep 2025

Sector-Level Displacement Estimates

Not all jobs face the same risk. Finance, administration, and legal services are most exposed — construction, healthcare, and hands-on care work are least. The difference is whether your job breaks down into tasks a language model can replicate.

The charts below use Government AI Occupational Exposure (AIOE) scores applied to 22,000 UK tasks. Figures scale in real time with the sliders on the Model tab — so what you see reflects your assumptions, not ours.

Jobs at Risk by Sector — Your Model vs IPPR/DSIT Baseline
Blue = IPPR/DSIT published baseline. Lime = your model output. Numbers in thousands.
AI Exposure by Occupation (AIOE Score)
Gov.UK / DSIT Nov 2023: finance & insurance most exposed. ONS: admin/secretarial workers 43% believe AI risks their job.
Scaled Displacement — Your Parameters
Jobs at risk ('000s) per sector under current slider values.
Data sources for sector figures: Finance & insurance exposure from DSIT Impact of AI on UK Jobs (Nov 2023), which used the Felten–Raj–Seamans AI Occupational Exposure score mapped to UK SOC codes. The IPPR analysed 22,000 tasks across the entire UK economy. The ONS Business Insights survey (Sep 2025) found 27% of AI-adopting firms expect admin/clerical roles most impacted. The GOV.UK assessment (Jan 2026) confirmed UK job postings in high-exposure occupations fell 38% more than low-exposure roles.

Fiscal Analysis — The Transition Trough

The less visible risk is what job displacement costs the government. Every worker who stops paying income tax and National Insurance leaves a hole in the public finances — potentially tens of billions a year at peak disruption.

This tab maps the "fiscal trough" — the years between automation taking hold and new jobs generating enough tax revenue to compensate. The question: could a token levy fill that gap before the new roles arrive?

Income Tax + NI Revenue — Baseline vs AI Scenario
Baseline ~£440B/yr (HMRC 2024–25). Modelled erosion from labour displacement vs recovery from new jobs and productivity-linked wages.
Token Levy Revenue — Producer vs Consumer Split
Based on UK share of global AI API consumption (~5–7% of $700B capex market). Tiered by levy rate and SME exemption.
Net Fiscal Position — Combined Scenario (2025–2035)
Trough depth and recovery timing under current parameters. Shaded area = cumulative deficit before levy. Line = with levy in place.
UK fiscal baseline assumptions: Income tax + NI receipts c. £440B/yr (HMRC 2024–25 outturn). Labour income constitutes approximately 73% of the income tax base. Average effective rate on displaced workers assumed at 28%. UK AI token consumption estimated at 5–7% of global API market. Growth in taxable base assumed at 35%/yr (consistent with UK AI sector 58% company growth DSIT 2024). Key caveat: GDP productivity gains may partially offset income tax losses through corporate tax and VAT on higher consumption — this model conservatively focuses on the labour tax channel only.

Model Methodology

Built a model, show your working. This tab explains the economics behind everything you see — the academic theory, the equations, the UK data sources, and the caveats. Nothing is hidden.

The model is based on Acemoglu & Restrepo's (2019) task displacement framework — the same work cited in the IPPR's and the Government's own analysis.

The Economic Theory Behind the Model
Based on Daron Acemoglu & Pascual Restrepo (MIT, 2019–2022) — the framework cited by IPPR and UK Government analysts

Think of every job as a bundle of tasks: filing documents, writing reports, answering queries, diagnosing faults, negotiating contracts. AI doesn't replace jobs all at once — it takes over tasks within jobs, one by one.


Displacement effect — always reduces GDP: Every task moved from a person to AI is a task whose wages no longer get paid. Lower wage income → less spending → lower GDP.


Productivity effect — raises GDP: AI does the same tasks cheaper and faster. That cost saving flows partly as profit, partly as lower prices, partly as higher wages for workers whose jobs AI augments.


Reinstatement effect — can offset displacement: Automation also creates demand for new kinds of work only humans can currently do. The ρ (reinstatement rate) slider controls how quickly this happens.


"So-so automation" (Acemoglu & Restrepo, 2022): If productivity gains are modest but displacement is large, automation can reduce GDP overall. Use the model's sliders to test whether current AI looks "so-so" or genuinely transformative.

Model Equations
Simplified parameterisation used in this tool
// Displacement
D = workforce × α × δ
// Reinstatement
R = D × ρ × recycle_multiplier
// Net employment change
ΔE = R − D
// GDP impact (Acemoglu decomp)
ΔGDP = (λ−1) − α×δ×s_L + ρ×α×δ×s_L_new
// Fiscal gap (labour tax channel)
ΔTax = D × avg_wage × eff_tax_rate
// Token levy revenue
L = UK_token_market × levy × (1−sme_adj)
// Coverage ratio
C = L / ΔTax
s_L = labour share of income (~0.57 UK). avg_wage = £38,000 (ONS 2025). eff_tax_rate = 0.28. UK_token_market grows at 35%/yr from £8B (2026 est.).
UK Data Sources
ONS LFS (Feb 2026) — 34.24M employed, 75.0% employment rate 16–64
ONS BICS (Sep 2025) — 23% of UK businesses using AI; 10% replacing roles
DSIT AI Sector Study (2024) — 3,256 UK AI firms; 58% company growth YoY
GOV.UK AI Labour Impact (Jan 2026) — UK job postings −38% in high-exposure
IPPR (March 2024) — 22,000-task analysis; wave-1 11%, wave-2 59% exposure
PwC UK AI Report — England GDP +10.6% by 2030
DSIT AI Jobs & Training (Nov 2023) — AIOE scores by UK SOC 2010
Parliament POST (Feb 2026) — balanced review of AI employment evidence
Goldman Sachs (2025) — +15% labour productivity, +7% GDP
OBR Fiscal Outlook (Nov 2025) — baseline tax revenue projections
Model Limitations & Caveats

1. Causality not established: GOV.UK (Jan 2026) notes job posting declines are "consistent with AI contributing" but causality unproven.


2. Task ≠ Job: IPPR emphasises that automating a task within a job rarely eliminates the whole job. Most displacement is through reduced hiring, not firing.


3. Token market estimate uncertainty: No official UK data on token consumption volume. £8–12B estimate derived from UK share of global cloud spend. ±50% confidence interval.


4. General equilibrium effects: This partial-equilibrium model omits capital income tax gains, VAT uplift from productivity, and second-order multiplier effects.


5. Behavioural responses: Token tax may reduce AI adoption, offshore API consumption, or accelerate automation to offset cost — all would change model outputs.