Workforce Resilience & AI Transition
A composite index measuring the Isle of Man's capacity to absorb AI-driven labour displacement without the safety valves available to larger economies. Combines automation exposure, occupational diversity, skills transferability, and adoption speed into a single resilience score per occupation group and geographic area.
Data generated: 4 Apr 2026
What does “Income at Risk” mean?
This is the total annual salary bill for workers in occupations where AI can automate a significant share of tasks — weighted by each occupation's automation share. It is not a prediction of job losses. It represents the income currently flowing to work that AI could technically perform. At £700.1M, this represents roughly 30.4% of the island's estimated ~£2.3bn total labour income. Note: the IoM's ~£7.7bn GDP is heavily skewed by corporate profits (~70%) — only ~30% flows to workers as wages and salaries. This means labour disruption hits workers disproportionately hard relative to headline GDP figures. The actual outcome depends entirely on adoption speed (kappa) and the island's reallocation capacity.
Isle of Man Labour Market — The Starting Position
Before interpreting resilience scores, understand the economy these workers operate in. This is real IoM data from Government statistics, not modelled estimates.
Why this matters: The IoM has one of the highest labour force participation rates in the world (84.4%) and extremely low unemployment. This means there is almost no spare labour capacity. When AI displaces workers, they cannot be absorbed by an existing pool of unemployed — the economy must create entirely new roles, retrain at pace, or accept permanent exits from the workforce. The vacancy-to-unemployed ratio shows how tight the market already is: every displaced worker is one fewer available to fill existing unfilled positions.
Why This Matters
The China Shock literature (Autor, Dorn & Hanson, 2013) demonstrated that geographically concentrated labour markets suffer disproportionately from economic shocks. Regions where a narrow set of industries dominated saw slower recovery, persistent wage depression, and long-term social costs that dwarfed the efficiency gains from trade liberalisation.
The Isle of Man is the ultimate concentrated market. With ~42,000 workers on an island of 85,000, there is no inter-state migration safety valve. A worker displaced from financial services in Douglas cannot simply commute to a tech hub 50 miles away. The entire labour market is a single commuting zone — and one where a handful of sectors (finance, e-gaming, government, professional services) account for the vast majority of high-skilled employment.
Levy Yeyati & Sosa (2024) added a crucial insight: it is not the extent of AI adoption that determines labour market harm, but the speed. Their kappa ratio — the displacement rate divided by the reallocation rate — predicts outcomes better than any automation probability estimate. When displacement runs ahead of reallocation (kappa > 0.65), even modest automation shares generate severe dislocations.
This index attempts to measure the island's absorptive capacity: can the IoM economy redeploy workers fast enough to keep pace with AI-driven task displacement? The answer depends on occupational diversity (are there adjacent roles to move into?), skills transferability (can existing skills map to new occupations?), and the adoption speed ratio itself.
Adoption Speed Gauge
0.370Caution — displacement outpacing reallocation
Kappa = displacement rate / reallocation rate. When displacement runs ahead of the economy's ability to redeploy workers, the ratio climbs. The Levy Yeyati finding: adoption speed matters more than adoption extent for labour market outcomes.
Which Occupation Groups Are Most Vulnerable?
This chart ranks every major occupation group on the island by overall resilience to AI disruption. The lower the score, the more urgent the need for intervention.
So what? Groups in red (<40) face a combination of high automation exposure, limited occupational alternatives, and low skills transferability. These are the groups where retraining programmes should be designed now, before displacement peaks. Green (>65) groups can likely self-adjust through market mechanisms. Worker counts show the scale of each group — larger bars mean more people affected.
Why Is Each Group Vulnerable (or Resilient)?
The composite score above hides important differences. This breakdown shows which dimension is driving each group's vulnerability — because the right intervention depends on the right diagnosis.
Reading the bars: Automation = how much of the work AI can do (higher = less exposed). Diversity = are there many sub-specialisms to move between? Transferability = do the skills translate to AI-augmented roles? Adoption Speed = salary runway during retraining. A group with low automation but high diversity is structurally different from one with high automation but low diversity — the policy response for each should be different.
Is AI Displacement Outpacing the Island's Ability to Absorb It?
This is the single most important chart on this page. The kappa ratio measures whether AI-driven job displacement is happening faster than the economy can redeploy workers into new roles.
So what? When kappa stays below 0.36 (green zone), retraining and new job creation keep pace with automation — the transition is manageable. Between 0.36 and 0.65 (amber), the system is under strain and active intervention improves outcomes. Above 0.65 (red), Levy Yeyati's research predicts permanent labour force exits accelerate — workers don't retrain, they leave. The blue line shows weekly job postings — when kappa rises while job posts fall, the squeeze is real.
What Kind of Work Is Being Advertised?
Not all automation is displacement. This chart tracks whether the island's job market is shifting towards augmentation (AI assists humans) or replacement (AI does the work).
So what? A healthy AI transition shows the augmented (amber) band growing while routine (red) shrinks — work is changing, not disappearing. If routine shrinks faster than augmented grows, the human-only (green) band must absorb the difference or workers leave the market entirely. For policymakers: track whether UCM and training providers are building augmentation skills (AI copilot workflows, prompt engineering, AI-assisted analysis) rather than teaching skills that are already routine.
How Concentrated Is Employment? The China Shock Warning
The China Shock research found that the most devastated communities were those where a small number of employers dominated the local labour market. When those employers restructured, there was nowhere else to go.
So what? HHI measures employer concentration — higher = fewer employers dominate hiring. Groups flagged with !!! have 3 employers holding >60% of vacancies. If any one of those employers restructures around AI, the impact on that occupation group is severe and immediate. For policymakers: concentrated groups need diversification support (attracting new employers to the island) alongside retraining. Retraining alone won't help if there are no alternative employers to move to.
Where Is the Money? The Fiscal Case for Intervention
This chart shows the total annual salary bill at risk in each occupation group — calculated as: workers x median salary x automation share. It answers the Treasury question: “what is the fiscal cost of getting this wrong?”
So what? The largest bars represent the greatest fiscal exposure. If displacement in these groups outpaces reallocation, the income tax, NI, and consumer spending loss flows directly to the Treasury. For a small economy with no transfer union safety net (unlike EU member states), every pound of lost income is felt locally. Compare the top bar to the managed scenario savings in the scenario modelling below — that delta is the return on investment for proactive retraining policy.
The Priority Matrix: Where to Focus Resources
This single chart can structure an entire policy prioritisation meeting. It plots every occupation group on two dimensions: how exposed they are (x-axis) vs how ready they are to adapt (y-axis). Bubble size = worker count.
Reading the quadrants: Top-right (red) = high exposure AND low readiness — urgent intervention needed. Top-left (amber) = high exposure but good readiness — monitor closely. Bottom-right (blue) = low exposure but also low readiness — build capacity now while there's time. Bottom-left (green) = low exposure, high readiness — deprioritise. Focus retraining budgets on the red quadrant first, then amber.
The Cost of Inaction vs the Return on Intervention
Three scenarios over 10 years. The difference between the red bar (accelerated) and the green bar (managed) is the return on investment for proactive policy.
So what? The managed scenario assumes active intervention: retraining programmes, employer engagement, controlled adoption pacing. It typically reduces cumulative displacement by 40-60% compared to letting the market set the pace. The income impact column shows this in pounds — that's the fiscal argument for Treasury. Note: “workers displaced” is cumulative over 10 years, not annual.
Where Can Displaced Workers Actually Go?
Retraining programmes fail when they're designed in the abstract. This chart shows specific, data-backed transition pathways — which at-risk occupations can realistically move to which resilient ones, based on skills overlap.
Reading the bars: Longer bars = higher skills overlap = cheaper, faster retraining. Green = high feasibility (most skills transfer directly). Amber = medium (6-12 month upskilling needed). Red = low (significant retraining required). Worker counts show the scale of each pathway opportunity. For UCM and training providers: these are the specific from-to routes to build programmes around.
Key Insights
Auto-generated findings from the latest resilience computation. Each insight is derived directly from the data — not editorial opinion.
Elementary Occupations is high-risk
ResilienceComposite resilience index of 33/100 indicates significant vulnerability to AI-driven workforce disruption.
Affected workers: 3,685
Recommendation
Prioritise targeted reskilling programmes and transition pathways for the 3,685 workers in this group.
Sales and Customer Service Occupations is high-risk
ResilienceComposite resilience index of 38/100 indicates significant vulnerability to AI-driven workforce disruption.
Affected workers: 1,747
Recommendation
Prioritise targeted reskilling programmes and transition pathways for the 1,747 workers in this group.
High income exposure in Professional Occupations
IncomeEstimated £153.8M of annual income is at risk from automation in this group.
Affected workers: 7,801
Recommendation
Model fiscal impacts and develop income support mechanisms for transition periods.
High income exposure in Managers, Directors and Senior Officials
IncomeEstimated £235.8M of annual income is at risk from automation in this group.
Affected workers: 7,628
Recommendation
Model fiscal impacts and develop income support mechanisms for transition periods.
High income exposure in Associate Professional and Technical Occupations
IncomeEstimated £112.4M of annual income is at risk from automation in this group.
Affected workers: 6,230
Recommendation
Model fiscal impacts and develop income support mechanisms for transition periods.
High income exposure in Administrative and Secretarial Occupations
IncomeEstimated £70.8M of annual income is at risk from automation in this group.
Affected workers: 6,060
Recommendation
Model fiscal impacts and develop income support mechanisms for transition periods.
Adoption speed ratio (kappa) at 0.37
Adoption SpeedCurrent kappa of 0.37 indicates displacement is outpacing reallocation capacity.
Affected workers: 43,308
Recommendation
Increase investment in retraining programmes and consider adoption-pacing policies to allow workforce adjustment.
Most resilient groups: Skilled Trades Occupations, Professional Occupations, Managers, Directors and Senior Officials
StrengthsThese groups have composite scores of 62, 56, 54 respectively, indicating strong adaptive capacity.
Affected workers: 20,345
Recommendation
Leverage these groups as anchors for workforce transition, using their skills transferability for cross-sector retraining.
Current transition is augmentation-led
Transition CharacterRoutine (automatable) tasks at 22.85% vs augmented tasks at 47.15% of vacancy task profiles.
Affected workers: 43,308
Recommendation
Continue supporting augmentation-led transition while monitoring for displacement acceleration.
Early Action Interventions
Priority-ranked interventions derived from the resilience analysis. These represent data-informed starting points for policy discussion — not prescriptions.
Launch targeted reskilling programmes for high-risk occupational groups
2 SOC major group(s) have composite resilience below 40, indicating urgent vulnerability.
Model fiscal impacts and design income transition support
Groups with over £50M income at risk require proactive fiscal planning to avoid economic shock.
Implement adoption-pacing policies and expand retraining capacity
Kappa of 0.37 indicates displacement is outpacing reallocation.
Scenario Modelling
Three adoption-speed scenarios over a 10-year horizon, based on logistic displacement curves calibrated to current kappa. The “managed” scenario assumes active policy intervention halves the displacement-to-reallocation ratio.
Baseline (status quo)
Accelerated AI Adoption
Managed Transition
Note: Displacement uses a logistic adoption curve (S-curve) with Levy Yeyati-calibrated parameters. “Workers displaced” represents cumulative 10-year displacement estimate, not annual. Income impact = displaced workers x weighted average salary x automation share.
Sector Transition Pathways
For at-risk occupation groups (composite < 50), the most viable transition targets based on skills overlap and salary proximity. Higher shared skills % = more feasible lateral movement.
- •Local government administrative occupations
- •Bookkeepers, payroll managers and wages clerks
- •Receptionists
- •National government administrative occupations
- •Chefs
- •Gardeners and landscape gardeners
- •Stonemasons and related trades
- •Roofers, roof tilers and slaters
- •Local government administrative occupations
- •Bookkeepers, payroll managers and wages clerks
- •Receptionists
- •National government administrative occupations
- •Chartered and certified accountants
- •IT quality and testing professionals
- •Primary education teaching professionals
- •Midwifery nurses
- •Exam invigilators
- •Shelf fillers
- •Postal workers, mail sorters and messengers
- •Farm workers
- •Chartered and certified accountants
- •IT quality and testing professionals
- •Primary education teaching professionals
- •Midwifery nurses
- •Exam invigilators
- •Shelf fillers
- •Postal workers, mail sorters and messengers
- •Farm workers
- •Financial accounts managers
- •IT user support technicians
- •Police officers (sergeant and below)
- •Science, engineering, and production technicians n.e.c.
- •Hairdressers and barbers
- •Pest control officers
- •Nursing auxiliaries and assistants
- •Sports and leisure assistants
- •Chartered and certified accountants
- •IT quality and testing professionals
- •Primary education teaching professionals
- •Midwifery nurses
- •Hairdressers and barbers
- •Pest control officers
- •Nursing auxiliaries and assistants
- •Sports and leisure assistants
- •Financial accounts managers
- •IT user support technicians
- •Police officers (sergeant and below)
- •Science, engineering, and production technicians n.e.c.
- •Sales and retail assistants
- •Customer service occupations n.e.c.
- •Call and contact centre occupations
- •Collector salespersons and credit agents
- •Chefs
- •Gardeners and landscape gardeners
- •Stonemasons and related trades
- •Roofers, roof tilers and slaters
- •Sales and retail assistants
- •Customer service occupations n.e.c.
- •Call and contact centre occupations
- •Collector salespersons and credit agents
- •Chartered and certified accountants
- •IT quality and testing professionals
- •Primary education teaching professionals
- •Midwifery nurses
- •Delivery drivers and couriers
- •Large goods vehicle drivers
- •Food, drink and tobacco process operatives
- •Taxi and cab drivers and chauffeurs
- •Chartered and certified accountants
- •IT quality and testing professionals
- •Primary education teaching professionals
- •Midwifery nurses
- •Delivery drivers and couriers
- •Large goods vehicle drivers
- •Food, drink and tobacco process operatives
- •Taxi and cab drivers and chauffeurs
- •Chefs
- •Gardeners and landscape gardeners
- •Stonemasons and related trades
- •Roofers, roof tilers and slaters
📐Methodology
The Composite Resilience Index (0-100) is an equally weighted average of four sub-indices, each scored 0-100:
1. Automation Exposure (inverted)
Based on Frey & Osborne (2017) automation probabilities and the Felten, Raj & Seamans (2023) AI Occupational Exposure Index. Higher score = less exposed. SOC-level probabilities are aggregated to major groups weighted by worker count.
2. Occupational Diversity
Inverse HHI of employer concentration within each SOC major group on the island. Adapted from Autor et al (2013) commuting-zone approach. Higher diversity = more employers = more reallocation options if any single firm automates.
3. Skills Transferability
O*NET skill overlap analysis — how many adjacent occupations share 70%+ of required skills with each SOC major group? Based on task-skill crosswalk methodology. Higher score = more lateral movement options for displaced workers.
4. Adoption Speed Readiness
Levy Yeyati kappa-derived measure of how well the island's retraining and vacancy creation pace matches observed automation adoption rates. Combines job centre vacancy flow, training programme throughput, and sectoral churn data.
All sub-indices are normalised to 0-100 before averaging. Missing data for any sub-index reduces the denominator (i.e., partial scores are still valid). Area-level resilience uses the same formula weighted by occupational composition of each area from census returns.
Academic References
Autor, D., Dorn, D. & Hanson, G. (2013). “The China Syndrome: Local Labor Market Effects of Import Competition in the United States.” American Economic Review, 103(6), 2121-2168. doi:10.1257/aer.103.6.2121
Levy Yeyati, E. & Sosa, L. (2024). “The Adoption Speed Dilemma: Technology Transitions and Labour Market Absorption.” CID Working Paper Series. Harvard Growth Lab
Frey, C.B. & Osborne, M.A. (2017). “The future of employment: How susceptible are jobs to computerisation?” Technological Forecasting and Social Change, 114, 254-280. doi:10.1016/j.techfore.2016.08.019
Felten, E., Raj, M. & Seamans, R. (2023). “Occupational, industry, and geographic exposure to artificial intelligence: A novel dataset and its potential uses.” Strategic Management Journal, 44(7), 1643-1681. doi:10.1002/smj.3286
ℹ️About this data
Census data: Isle of Man Census 2021 — occupation counts by SOC major group and area.
Automation probabilities: Frey & Osborne (2017) base probabilities, updated with Felten, Raj & Seamans (2023) AI exposure indices. Cross-walked from O*NET SOC to UK SOC2020 via the ONS crosswalk tables.
Salary data: IoM Income Tax Division anonymised salary bands cross-referenced with ONS ASHE for comparable UK SOC groups where IoM-specific data is insufficient.
Vacancy flow: Weekly job postings from IoM Job Centre (Locate.im), covering advertised positions since 2024.
Kappa calculation: Rolling 12-week displacement estimate (job losses + automation announcements) divided by rolling 12-week reallocation estimate (new hires + training completions + vacancy creation rate).
Updated weekly when new vacancy data is available.
Analysis by Smart Island / Manx Technology Group. Census data: IoM Government. Automation indices: Frey & Osborne (2017), Felten et al (2023). Adoption speed framework: Levy Yeyati & Sosa (2024).