AI will transform 40% of jobs. Twelve million occupational transitions will be needed across Europe by 2030. The disruption will be comparable to the Industrial Revolution. The framing is everywhere: unprecedented change is coming, and we must prepare for a new era of work instability.
But this framing is wrong.
Not because the projections are false. They may well prove accurate. But because they misdiagnose what’s actually new. Work instability isn’t coming. It’s been here all along. What AI threatens isn’t disruption itself.
It’s synchronisation: the same disruption hitting everyone at once.
This distinction matters because it determines what we build. If AI creates a new problem, we might reach for crisis tools: taskforces, emergency programmes, time-limited responses to a temporary shock. If AI intensifies a permanent condition, we need permanent infrastructure.
A field hospital or a health service. We keep building the former when we need the latter.
The stability that never was
The “job for life” is invoked as historical baseline and lamented loss. But for most workers, it was never reality.
Women moved in and out of paid employment around caring responsibilities, not as brief departures from stable careers, but as part of the normal pattern of working life. Seasonal labourers followed harvests and construction cycles. Service workers moved between roles as businesses opened and closed. Workers in declining industries such as textiles, mining, shipbuilding, and steel lived with redundancy as a recurring threat, not a one-off shock.
The stable career was real, but specific. It belonged to particular sectors, institutions, and decades: unionised manufacturing, public employment, large firms with internal labour markets and defined-benefit pensions, and professions with structured progression and institutional continuity.
For a substantial minority of workers in specific decades, stability was genuine experience. For the rest, instability was simply what work meant.
The problem was that this minority experience became the policy assumption for everyone. Career guidance assumed you would choose a path and follow it. Pension systems assumed continuous employment. Benefits systems assumed unemployment was a temporary interruption to an otherwise stable working life.
Those whose lives fitted that model were Churchill’s “genuine working men.” Those whose working lives were fragmented, interrupted, or repeatedly reshaped were treated as exceptions, failures, or suspect. Chaotic work histories were read as personal deficiency rather than structural reality.
We built policy around the minority experience and pathologised the majority’s.
I see the hangover of this every day in my work at Stay Nimble. When I hear the stories of people who have navigated multiple career changes, periods out of work, sideways moves, or false starts, the system has conditioned them to view this as a personal failure. But it isn’t failure. It’s simply what a working life looks like for most people. The system treats reality as an aberration.
What endemic means
Health systems distinguish endemic from epidemic disease. Endemic conditions are permanently present in a population, always generating some demand for treatment and requiring standing capacity to address. Epidemic conditions surge beyond a baseline: sudden spikes that overwhelm normal capacity and require an emergency response.
Career transitions are endemic.
In any given year, some percentage of workers will face redundancy, career change, return from caring responsibilities, health-related work capacity changes, industry decline, or simple dissatisfaction requiring a new direction. This has always been true. It will always be true. The rate may vary, but the phenomenon is permanent.
Endemic conditions require infrastructure. You don’t build a hospital when disease arrives; you maintain one because disease is always present. You don’t train doctors in response to illness; you sustain a medical profession because illness is the permanent condition of any population.
Career support has never received this logic.
We build programmes when unemployment spikes. We fund interventions when political pressure demands response. We treat transitions as problems to be solved and then forgotten, rather than permanent features of working life requiring permanent capacity to address.
The result is predictable. When people need support, it may or may not exist depending on what programme is currently running, what eligibility criteria apply, and whether funding has been renewed. The field hospital appears when crisis is undeniable, then disappears when attention moves elsewhere.
Meanwhile, the endemic need continues, unmet, invisible, assumed to be someone else’s problem.
I’ve watched this cycle repeat throughout my career. A programme launches with fanfare. People start engaging with it. They begin to trust it. And then the funding ends, or the contract changes, or the eligibility criteria shift, and they’re left with nothing, sometimes at exactly the moment when the support was starting to make a difference.
That’s not a system. That’s a series of disappointments dressed up as policy.
What epidemic means
Epidemic conditions overwhelm endemic capacity. The NHS just about handles seasonal flu through normal operations. It nearly collapsed under COVID because demand synchronised: everyone sick at once, everyone needing care simultaneously, the queue suddenly infinite.
AI threatens epidemic conditions for career support. Not because it creates instability. Instability was always there. But because it might synchronise instability across sectors, geographies, and timescales in ways that distributed historical disruption never did.
Consider the pattern of previous technological displacement. Manufacturing automation unfolded over decades. Call centre offshoring happened sector by sector. Retail’s shift online accumulated gradually. At any given moment, specific workers in specific industries faced disruption, but the broader system could absorb them through natural turnover, geographic mobility, and adjacent opportunities.
Now imagine software engineering, administrative work, customer service, junior professional roles, creative tasks, and logistics planning all contracting within a five-to-ten year window. The sectors are different. The skills are different. The affected populations are distributed across the entire economy. But the timing converges.
This is the epidemic scenario: not unprecedented instability, but unprecedented synchronisation of instability. Everyone needing support at once. The queue infinite. Whatever endemic capacity exists (and currently, there is almost none), would be overwhelmed instantly.
We don’t know if this scenario will materialise. AI projections range from modest labour market adjustment to fundamental economic restructuring. The IMF’s 40% figure describes jobs “exposed to” AI, not jobs eliminated. Many roles will be augmented rather than replaced. The timeline is uncertain. The distribution is unclear.
But the possibility of synchronised disruption is what makes the current moment different from previous waves of technological change.
The asymmetry of risk
We don’t know the future. But the asymmetry of risk is clear, and this is what I keep coming back to.
If we build endemic infrastructure and the epidemic doesn’t arrive, we’ll have career support that should have existed anyway, serving the permanent reality of work transitions that has always been present and always been underserved. The investment isn’t wasted. It addresses a need that exists regardless of AI.
If we don’t build infrastructure and the epidemic does arrive, we face synchronised demand on a system that doesn’t exist. Emergency programmes will be hastily commissioned. Providers will lack established relationships and local knowledge. The Restart Scheme precedent of building capacity during crisis rather than before it will repeat at a larger scale with worse results.
The pandemic offered a preview. Furlough protected jobs, but when it ended, employment services faced demand they couldn’t meet. Restart was commissioned rapidly, primes scrambled to build delivery capacity, and early performance was poor. The NAO documented providers building capability in real time while simultaneously expected to deliver results.
The system had no resilience because there was no system, only programmes that could be spun up when crisis forced action.
What we’re talking about is career support infrastructure that should exist for endemic conditions, stress-tested by epidemic possibility. The AI scenario makes the case urgent. But the case itself doesn’t depend on AI. It depends on finally acknowledging that work transitions are permanent, not exceptional, and building accordingly.
What surge capacity means
Health systems maintain baseline capacity for endemic demand and surge capacity for epidemic conditions. Beds can be added. Staff can be redeployed. Elective procedures can be postponed. The system stretches, imperfectly and sometimes catastrophically, but the capacity to stretch exists because someone designed for it.
Career support has no equivalent.
There is no baseline capacity to stretch from. There is no surge mechanism. There are programmes that exist or don’t, funding cycles that align or don’t, eligibility criteria that include or exclude.
Surge capacity for career support would require several things we don’t currently have.
Standing infrastructure to surge from. You cannot scale what doesn’t exist. Universal access to career support, the endemic baseline, is a prerequisite to epidemic response. Trying to build capacity during a crisis means starting from zero every time.
Flexible delivery mechanisms. Human coaching doesn’t scale infinitely. AI-augmented support can extend reach when demand exceeds human capacity, but only if it’s already integrated, already trusted, already part of how people access help. Bolting on technology during crisis fails for the same reasons emergency procurement always fails: no relationships, no learning curve, no embedded practice.
This is part of why we’ve been building Ask.Nim the way we have. Not as a crisis response tool, but as part of normal infrastructure, so that if demand surges, there’s something already working that can stretch.
Modular intensity. Some people need intensive support. Others need light-touch guidance. Surge conditions require shifting resources toward triage: identifying who needs most help, providing enough to everyone, concentrating capacity where it matters most. This requires systems designed for variable intensity, not programmes with fixed engagement models.
Cross-sector coordination. Epidemic career disruption won’t respect departmental boundaries. Health services will see increased demand from work-related stress and displacement-linked conditions. Housing services will see pressure from income disruption. Benefits systems will face claim surges. Effective response requires integration across these domains; the same integration that IPS research identifies as crucial for endemic support.
None of this is possible without first building for the endemic condition. Surge capacity assumes something to surge. We have nothing.
Why we’re not preparing
The pandemic revealed how societies respond to epidemic risk they’d previously ignored. Decades of warnings about pandemic preparedness went unheeded. When crisis arrived, the absence of preparation was catastrophic. Then, briefly, preparation became political priority. Now, attention has moved elsewhere. The next pandemic will likely find us similarly unprepared.
AI labour market disruption faces the same dynamic. Warnings exist. Projections are published. Think tanks issue reports. Nothing happens because the crisis isn’t yet acute, and preparing for non-acute crisis has no political constituency.
The Vaccine Taskforce was possible because people were dying. The political cost of inaction was immediate and visible. AI displacement hasn’t produced equivalent pressure. Unemployment statistics tick up gradually. Individual stories of displacement don’t aggregate into undeniable emergency. Ministers can defer action without facing consequences today.
There are two ways to approach this.
The first is crisis opportunism: wait for disruption to become undeniable, then build rapidly in the window when political will exists. This is how most policy happens. It’s also why most policy is inadequate; because capacity built during crisis is worse than capacity built before it.
The second is the endemic argument: stop claiming that AI creates unprecedented threat, and start arguing that we’ve never built for the permanent reality of work transitions. The case for infrastructure doesn’t depend on AI predictions proving accurate. It depends on recognising that career instability has always been present and has never been properly addressed.
I understand why the endemic argument is harder to make politically. It lacks urgency. It doesn’t generate headlines. But it’s the honest one, and it’s the one I keep coming back to, because it builds something that works regardless of whether the epidemic arrives.
What we’re actually deciding
The choice isn’t whether to respond to AI. Governments will respond when disruption becomes visible. Programmes will be announced. Funding will be allocated. The question is whether that response builds on existing infrastructure or starts from nothing.
If we build endemic capacity now, with universal career support, proper funding, fidelity to evidence-based practice, and integration across health, housing, and benefits, then epidemic response has something to work with. Surge mechanisms can activate. Existing relationships can intensify. Proven approaches can scale. The crisis will still be hard. It won’t be starting from zero.
If we don’t build endemic capacity, we’ll face epidemic demand on a system that doesn’t exist. Emergency programmes will repeat the failures of every previous emergency programme. Providers will lack capability. Integration will be absent. Outcomes will be poor. And when the acute phase passes, the programmes will wind down, leaving nothing permanent, until the next crisis forces us to start over.
The pandemic should have taught us that preparing for predictable crises is cheaper than responding to them. We’re not acting on that lesson for career support. We’re not acting on it despite knowing that the endemic need is already unmet, that the epidemic risk is plausible, and that building now serves both conditions.
We keep waiting for a crisis to force action. The alternative is to build for the reality we already have, and ensure that whatever’s coming finds something other than nothing.
What we’re doing anyway
I’ve spent six pieces describing a system that doesn’t work and an alternative that could. I’m aware that describing problems is easier than solving them, and that the forces holding the current system in place are substantial.
But I’m not waiting for permission to build something different.
Stay Nimble exists because I think career support should be infrastructure, not programmes. We’re small. We can’t fix the system. But we can build something that works the way we think it should: sustained relationships, integration with other support, time-unlimited availability, fidelity to what the evidence says matters. And demonstrate that it’s possible.
We’re connecting thousands of people with registered career development professionals through our platform. All ages, all stages. Neurotypical, neurodiverse. Chronically ill, fighting fit. We’re building Ask.Nim because AI can extend what those human coaches do, not replace it. Because endemic infrastructure needs flexible delivery mechanisms. Because if demand surges, we want something already working that can stretch.
We’re supporting communities, volunteers, and frontline workers to deepen their skills in effective career conversations through the Career Allies microcredential. Building nationwide communities of practice that stand beyond programmatic churn. We’re measuring fidelity because outcomes depend on factors we can’t control, and what we can control is whether we deliver evidence-based practice properly. We’re trying to be permanent in a landscape of programmes. Trying to build relationships that don’t end when funding cycles turn over. Trying to prove that the alternative isn’t just theoretically better but practically achievable.
I don’t know if the epidemic is coming. I don’t know if AI will synchronise disruption or if the transition will be gradual enough to absorb. I don’t know if governments will ever build the endemic infrastructure that should have existed for decades.
But I know the endemic need is real. I see it every day. And I know that building for it, properly and permanently, is worth doing regardless of what happens next.
A system for a permanent condition
AI isn’t creating work instability. Work instability is endemic: permanent, distributed, present in every year and every economy, affecting more people than the mythology of stable careers ever acknowledged.
What AI might create is epidemic instability: synchronised disruption across sectors and timescales, overwhelming whatever capacity exists. The threat isn’t the disease. It’s the surge.
We have no endemic capacity to surge from. We have programmes instead of infrastructure, eligibility criteria instead of universal access, outcome measurement instead of quality assurance, two-year funding cycles instead of permanent provision.
The case for building isn’t about AI. It’s about finally acknowledging what has always been true: career transitions are normal, support is needed throughout working life, and we have never built for this reality. AI just makes the cost of continued failure potentially catastrophic rather than merely unjust.
Endemic infrastructure with epidemic surge capacity. Permanent provision that can scale. Not a taskforce for a crisis, but a system for a permanent condition.
We could build it. We could have built it any time in the last 117 years. The question is whether we will.
I’m not waiting to find out. There’s work to do now, for the people who need support now, with whatever we can build now. That’s where I’m putting my energy. The system might never change. But what we build doesn’t have to wait for it.
Note: I have used Perplexity to help with researching and editing this series.