Which European cities will survive the AI disruption — and which won't?

AI is not a future threat. For millions of European workers in finance, administration, logistics, and knowledge work, displacement is already beginning. The question isn't whether your city will be affected — it's whether it has the infrastructure, culture, and resources to help its people adapt.

We spent considerable time building a ranking that goes beyond the obvious. Most analyses look at which jobs are automatable. We asked a harder question: when disruption arrives, which cities give their workers the best chance of surviving and rebuilding?

The answer depends on nine factors — and crucially, how you weight them.

The framework

We scored ten of Europe's largest cities (1 million+ inhabitants) across nine dimensions. The scores reflect both quantitative data — VC investment volumes, broadband speeds, business formation rates, retraining programme uptake — and qualitative assessment of cultural and institutional factors that data alone doesn't capture.

What we found

The cities that score highest on traditional resilience metrics — Zurich, Amsterdam, Stockholm — are not necessarily the cities where an ordinary displaced worker is best placed. There is a persistent and important gap between city economic strength and worker outcome quality.

The most counterintuitive finding: under realistic crisis conditions — where government programmes lag and private capital moves first — Barcelona emerges as one of the more resilient cities for ordinary workers. Its 9/10 affordability score means a displaced Barcelonan can survive two to three times longer on savings than an equivalent worker in Oslo or Zurich while the ecosystem catches up. And its job resilience score — 8/10 — is among the highest in the ranking.

The automation-resistance paradox

There is a persistent and damaging assumption baked into most AI resilience analysis: that high-wage cities are resilient cities. They are not. A city full of lawyers, bankers, and management consultants is significantly more exposed to AI than a city full of plumbers, chefs, and hotel workers.

This is counterintuitive because we associate professional services with stability and security. But AI excels at exactly the things those jobs require: pattern recognition in large datasets, drafting structured documents, processing contracts, analysing financial statements, routing customer queries. A paralegal producing first drafts, an analyst building Excel models, a customer service manager handling standard complaints — these roles are already being reduced by AI tools deployed today.

The jobs AI struggles with most are almost the opposite of prestigious. Physical presence in unpredictable environments. Fine motor dexterity. Reading the emotional state of a stranger in a hotel lobby. Improvising when a pipe runs differently than the plans suggested. Caring for an elderly patient who needs reassurance as much as medication. These require embodied intelligence that no current AI system can replicate.

This is why Barcelona and Milan score 8/10 on job resilience — among the highest in the ranking — despite being bottom-tier on retraining and funding. Their economies are built around tourism, hospitality, artisan manufacturing, and craft trades. These are genuinely hard sectors to automate at scale. Munich and Paris score lower despite higher average wages, because their large automotive and professional services workforces are facing structural disruption from both AI and the electric vehicle transition simultaneously.

London presents the starkest version of this paradox. It scores 7/10 on job resilience — solid, but not exceptional — because while it has a large service and hospitality economy, it also has an enormous concentration of finance, legal, and administrative roles that are precisely the targets of the current wave of AI deployment. A city of extremes: world-class automation-resistance in some sectors, world-class automation-exposure in others.

City snapshots

The tool: weight what matters to you

Equal weighting across nine dimensions isn't honest. A 55-year-old factory worker made redundant in Munich needs different things from a 28-year-old software developer considering a move from London. The tool below lets you weight the dimensions to reflect your own situation — or your own view of what matters most when disruption hits.

The default weights reflect a realistic AI disruption crisis scenario: private capital and entrepreneurship move faster than government programmes; cost of living determines how long workers can survive while the ecosystem adapts; language and digital infrastructure unlock global remote markets. Retraining is set to 1 (minimum) not because it doesn't matter, but because government programmes typically take 2–5 years to design and scale — too slow for the first wave of displaced workers.

Interactive resilience ranking

Set the importance of each dimension from 1 (not important) to 10 (critical). City scores are recalculated as a weighted average out of 10. Use the preset buttons to explore different worker profiles.

All dimensions scored 1–10. Cost of living is scored inversely — a higher score means more affordable. Job resilience is scored on automation-resistance, not wage level — see note below the table.

Note on job resilience scoring: This dimension measures how automation-resistant the existing workforce is — not how high-wage or prestigious the jobs are. Tourism, hospitality, skilled trades, healthcare, and artisan work score highly because they require physical presence, human judgment, and dexterity that AI cannot replicate. Finance, law, and admin score lower despite higher wages, because these are precisely the roles AI is already beginning to replace. Barcelona and Milan score 8 because their economies are dominated by tourism, hospitality, and craft manufacturing. London and Paris score lower because their large finance and professional services workforces are significantly exposed.

The north-south divide, made visible

Whatever weighting you apply, a pattern emerges: northern and central European cities dominate the top half, southern European cities cluster at the bottom. This isn't coincidence — it reflects decades of divergence in social infrastructure investment, vocational training culture, and institutional capacity.

The partial exception is Barcelona, which punches above its weight on affordability and entrepreneurship. If the EU's structural funds are deployed effectively and Spain's retraining infrastructure improves, Barcelona has the raw ingredients to close the gap significantly over the next decade.

What London's paradox tells us

London is perhaps the most instructive case. It scores 10/10 on entrepreneurship and 9/10 on funding — unmatched in Europe on both. It has DeepMind, the largest VC market on the continent, and more AI investment than any city outside San Francisco and New York.

Its job resilience score sits at 7/10 — solid but not exceptional. London has a genuine paradox at its core: an enormous hospitality, trades, and service economy that is genuinely hard to automate sits alongside one of the largest concentrations of finance, legal, and administrative work in the world, which is highly exposed. The city contains both the most resilient and the most vulnerable workers in Europe, often in the same borough.

Yet under most "ordinary worker" weightings, London sits mid-table. A decade of austerity cuts to adult education, no legal retraining entitlement, Brexit's removal of EU work rights, and a cost of living that leaves displaced workers almost no runway — these structural weaknesses mean that London's extraordinary private sector strength is largely inaccessible to the workers who need it most.

The city is excellent for ambitious self-starters with capital and connections. It is poorly equipped for the median worker facing AI displacement.

The crisis default: why retraining is weighted low

The default importance setting for government retraining infrastructure is 1 out of 10. This is intentional and deserves explanation.

Government retraining programmes are built for gradual labour market shifts. They require legislation, funding cycles, procurement, and years of design and iteration. The UK's Lifetime Skills Guarantee took three years to launch and was immediately underfunded. Germany's Kurzarbeit scheme works well precisely because it was built during a previous crisis and refined over decades.

In a rapid AI disruption scenario — the kind that displaces 10–20% of knowledge workers within 3–5 years — the first wave of affected workers will not be caught by programmes that don't yet exist. Private bootcamps, employer-funded schemes, peer networks, and entrepreneur-created training businesses will fill the gap first. The "Govt safety net" preset shows what the ranking looks like once those programmes do exist — and it's a very different picture.

What this means if you're thinking about where to live and work

If you're a knowledge worker concerned about AI displacement, the honest answer is that geography matters enormously — and most people underweight it when making location decisions.

The factors that matter most in the next five years are not the ones that made cities attractive in the last twenty. Proximity to a major corporate employer matters less when that employer may automate your role. Proximity to a dense entrepreneurial ecosystem, affordable survival runway, and the legal right to work across multiple countries matters more.

By those measures, Amsterdam and Stockholm are the most strategically positioned large European cities. Berlin offers a compelling combination of startup depth, German retraining infrastructure, and relative affordability. Barcelona is significantly underrated for workers who prioritise survival time over institutional support.

London remains the highest-ceiling, highest-risk option — exceptional for those who can navigate its private sector, bruising for those who can't.

Analysis developed through iterative research and modelling at inagentic.ai. Scores represent assessments as of mid-2026 and combine quantitative data sources with qualitative judgement. We welcome challenges to specific scores.