AI Strategy Scotland: The Economics of New Jobs and What Business Leaders Need to Know

The Economics of New Jobs and What Business Leaders Need to Know

Most of the analysis being published about AI and employment is answering the wrong question. Supply has eaten the conversation: which roles are most exposed, which tasks can be automated, which professions sit on safer ground. All of it rests on an assumption that the amount of work to be done stays roughly constant as the cost of doing it falls.

That assumption has been wrong every time it's been properly tested. In textiles, in print, in manufacturing, in financial services. And there's nothing obvious about the current wave of AI adoption that suggests this time is any different.

The World Economic Forum's 2025 Future of Jobs Report, drawing on data from over a thousand global employers, projects 92 million roles displaced by 2030 and 170 million created, a net gain of 78 million (WEF, 2025). PwC's 2025 Global AI Jobs Barometer, which analysed close to a billion job ads across six continents, found employment growing even in the most automatable roles. Wages in AI-exposed industries are rising at roughly twice the rate of less-exposed sectors (PwC, 2025). The numbers don't support the doom narrative. They point to something more interesting happening underneath it.

The Historical Pattern Worth Taking Seriously

In the 1810s and 1820s, the Luddite movement across the East Midlands and West Riding was animated by a straightforward reading of supply economics. Power looms were coming. Weavers would become surplus. The reading was correct about the short run and wrong about everything that followed. Cheaper cloth didn't produce a world with the same quantity of cloth made by fewer people. It activated demand for cloth that hadn't previously existed at scale, opening markets, creating applications, and bringing in buyers who couldn't previously access the product at all. Employment in the textile industry grew across the following decades. The handloom weavers displaced in the transition didn't benefit from that growth, which matters and will come back to, but the direction of aggregate employment moved opposite to what the pessimists had mapped out.

Desktop publishing in the 1980s tells a similar story, though the timescale compressed dramatically. Print unions anticipated job losses from digital typesetting and they weren't wrong about the short run. What followed was a surge in print output across categories that had previously been too expensive to produce at all: short-run catalogues, newsletters, local marketing materials, self-published work. Graphic design headcount grew. The Wapping dispute was real and the disruption cost people dearly, but the sector as a whole didn't contract.

Economists have a name for the error threading through both failed predictions. The lump of labour fallacy: the idea that there's a fixed quantity of work available, so any gain in productive efficiency just takes jobs out of the system. It doesn't work that way. Never has. The fallacy has failed as a predictive model enough times, across enough different industries and centuries and countries with very different labour market structures, that using it as the default frame for the AI employment debate needs some justification. Which it isn't getting.

Why Demand Doesn't Stay Still

The reason the lump of labour fallacy keeps failing is demand elasticity. Services don't have a natural fixed quantity. Demand for them stretches and contracts in response to price, availability, complexity, continuity, personalisation, and the degree to which a service feels human rather than mechanical. AI is pulling on most of those dimensions at once. Any AI strategy Scotland businesses are developing in 2026 needs to account for demand-side expansion, not just displacement risk on the supply side.

Price is the obvious entry point. There are roughly 5.5 million small businesses in the UK, and the vast majority of them operate without professional marketing support, without meaningful legal resource, without financial planning beyond basic compliance, without strategic advice of any kind. This isn't for want of need. It's because the cost structures of professional services have never been designed around a ten-person business in Burnley or a three-person accountancy in Paisley. A retained agency at £3,000 a month. A solicitor charging £250 before the context is even established. The professional services market has always contained an enormous untouched base of organisations for whom the price never made sense. When AI assistance brings the cost into range, that base activates. Understanding where most UK small businesses currently sit in the AI adoption curve makes clear how significant that activation is likely to be.

PwC's 2025 data quantifies the effect. Between 2018 and 2024, productivity per worker in AI-exposed industries grew by roughly 27% against about 9% in less-exposed sectors (PwC, 2025). That kind of productivity gain doesn't stay inside the firm. It transfers into the price the firm can charge, and into the markets the firm can profitably reach. Markets that were previously dormant because the labour maths didn't work become viable to serve, and whole new entrants build business models around the new price floor.

Availability is a harder problem and the NHS makes it visible in ways private markets tend to obscure. Physiotherapy waiting lists measured in months across most of the country. ADHD assessments running to two years in parts of Scotland. Mental health provision that's functionally inaccessible until a crisis has already arrived. The barrier here isn't primarily cost. It's scarcity, and scarcity breeds futility: people stop seeking what they've concluded they won't get. When AI-assisted monitoring makes continuous contact with the health system plausible rather than aspirational, latent demand resurfaces in volume. Clinical professionals don't get displaced by this. They redeploy toward cases that actually need them, while the total number of people engaging with care increases substantially.

Something else is operating in parallel that gets less attention. A large proportion of the UK population is effectively shut out of services they're entitled to or would materially benefit from, because the systems are too complex to navigate without expertise. Benefits entitlements. Planning applications. Employment rights. Pension decisions. The people who know how to work these systems extract significant value from that knowledge. Everyone else either goes without or pays for guidance they arguably shouldn't need. Making complex systems legible doesn't make the experts redundant. It creates demand for people who position themselves between the tool and the person using it, handling what the tool can't manage, absorbing the frustration and fear the tool can't acknowledge. That role isn't going to AI.

The category that produces genuinely new economic territory, rather than democratising access to existing services, is continuity. A large number of professional service relationships are episodic because continuous delivery was impossible to staff at any price that worked. Healthcare again: you present with symptoms, something happens, the relationship largely pauses until the next episode. The idea that a GP practice might maintain adaptive ongoing contact with a patient over time, monitoring indicators and flagging changes before they become acute, has been inconceivable within NHS budgets for the better part of a generation.

AI changes the cost structure of monitoring enough to make that model viable. It doesn't deliver clinical judgement or the accountability that makes continuous care trusted, but it makes those elements deployable in configurations that haven't existed before. Roles get created for people to occupy that haven't been needed, because the service requiring them wasn't available.

Personalisation sits adjacent to all of this. When AI makes generic services cheap and accessible, two things happen in the market at the same time. People who previously received nothing at all gain access to AI-personalised alternatives. People who could access the full human version gain a clearer sense of what they're actually paying for, and the premium on human relationship sharpens rather than erodes. Both generate economic activity. The cheap tier doesn't hollow out the premium one.

Relational and value elasticity is the most overlooked of the six and probably the most consequential over a longer horizon. As AI commoditises the functional output of a service, the things that don't transfer become more valuable to the market. Who delivered the work. The relationship that came with it. The provenance attached to it. The accumulated trust that informed it.

None of those things disappear when the technical capability becomes ubiquitous. They become more legible, not less. People don't just want a coffee. They want the cafe. They don't just want a property transaction concluded. They want a solicitor they trust to make the call when something doesn't look right. The relational sector grows precisely because AI makes the alternative ubiquitous and cheap.

The Two Unlocks: How AI Expands the Demand Frontier

If the six elasticity types describe how demand stretches, the practical question is which direction it stretches in. There are two distinct unlocks at work, and they produce different economic effects.

The Affordability Unlock is the simpler of the two. It takes an existing service that already had clear demand and lowers the price to a point where new buyers can finally access it. The product doesn't change. The accessible market does. Most of the small business example from earlier belongs here. Design work that used to cost £5,000 becomes accessible at £500. Legal review at a fraction of the previous cost. Marketing strategy that doesn't require a retained agency. Financial planning for households that were previously priced out of professional advice. The market for these services broadens substantially. People who were always going to want the service if they could afford it become buyers for the first time, and the providers serving them are AI-augmented professionals, not AI alone.

The Possibility Unlock works differently and opens up more interesting economic territory. It makes service models operationally viable that couldn't have existed before, at any price, because the underlying labour maths made them impossible. Continuous preventative healthcare. Always-on financial life support for ordinary households. Personalised education that adapts in real time to how a specific person actually learns. These aren't cheaper versions of existing services. They're whole new categories that the underlying economics previously made inconceivable. And because the category itself is new, the demand emerging around it is genuinely additional, not redistribution from somewhere else.

The distinction matters because the two unlocks create different kinds of jobs. The Affordability Unlock expands existing role categories to serve new market tiers. The Possibility Unlock creates role categories that don't yet exist in meaningful numbers, because the service models they support haven't been viable until now. Both are happening at once. Both create employment. The standard "jobs at risk from AI" framing catches neither, because that framing only looks at supply.

The Human Premium Is an Economic Category, Not a Preference

The objection that arrives here is that AI will eventually perform the new roles as capably as it performs the old ones, so the employment gains are temporary at best. Worth engaging with properly, because the assumption embedded in it is specific: that what people buy when they purchase a professional service is purely the functional output. AI can produce the output. The human becomes surplus.

This is where the analysis goes wrong.

Accumulated context might be the most underrated thing inside a long professional relationship. An accountant who has worked with a family business for eight years holds something that has no straightforward equivalent in a new engagement. She knows the decision the founders argued about for six months and got wrong. She knows whose approval matters most for anything touching capital. She knows what the business owner is and isn't willing to hear delivered directly. That context has real economic value. The switching cost of rebuilding it gets badly underestimated. AI-augmented professionals who can integrate and recall years of client history don't replace this dynamic. In some setups they deepen it.

Accountability works differently. When financial advice leads to a loss, when a care plan fails, when legal strategy backfires, someone needs to own the outcome. Not metaphorically own it, but actually face professional and sometimes legal consequences for having signed off. AI generates recommendations. It can't be held responsible for consequences in any form that regulation currently requires or that markets currently recognise. As AI takes on more analytical work in regulated sectors, the professionals involved shift toward oversight and sign-off roles that exist precisely because accountability requires a person at the end of the chain. That's a different job, not an absent one.

Physical presence carries more load across more roles than most forecasts capture. The physiotherapist watching how a patient walks across the room. The social worker doing the home visit where the real picture of a family's situation is legible in ways no digital interaction captures. The consultant who shows up to the board meeting and can read what isn't being said. Remote monitoring and AI assistance handle the routine and the predictable. What's left clusters around the complex, the frightening, and the cases where the stakes are high enough that someone needs a person in the room with them. And the more efficiently AI handles the routine, the more the residual skews in that direction.

Behaviour change is one of those categories that spans psychology and economics without sitting neatly in either, and it gets left out of employment forecasts almost entirely. Which is a strange omission given the volume of evidence behind it. Across health, financial planning, educational outcomes, and business transformation programmes, the finding is fairly consistent: adherence is higher when a human relationship accompanies the plan.

Social commitment, accountability to a specific person, and the responsive adjustment that comes from someone who notices when something isn't working and asks about it directly. For any service where compliance is a substantial part of the value being created, AI-only delivery weakens the proposition in ways that clients notice.

Translation will expand as a category, and current forecasts treat it as marginal when it isn't. Converting a client's half-formed intent into a well-specified instruction that an AI system can reliably execute is a real skill, unevenly distributed, and unlikely to become universal at any pace that tracks deployment of the tools themselves. The productivity gap between people who use AI effectively and people who don't is already visible in early adopter data. As the tools become more capable, the people who understand them gain proportionally more from them. The human translator between the client and the AI is a role that grows in value as the underlying systems improve. Most forecasts miss this because the obvious story runs the other way.

Provenance sounds soft and isn't. The premium attached to things made or advised upon by a specific known person has survived every previous wave of automation. Craft brewing didn't disappear when industrial lager got cheap. It clarified, and it expanded. Bespoke professional relationships get more precisely valued when AI-generated alternatives become abundant. The contrast sharpens. The human signature on the work becomes more legible, not less, when the alternative is widely available.

Harvard Business School research by Chen, Srinivasan and Zakerinia (2025) puts numbers behind this argument. Analysing US job postings, they found that demand and skill requirements dropped in structured cognitive-task roles where AI substitutes for human judgement, while job postings for human-AI collaboration roles rose by around 22% per quarter within firms that adopted generative AI. The pattern matches the human premium reading directly. Work requiring accountability, translation, judgement, and human relationship doesn't disappear. It grows in volume and in skill complexity.

The Human Premium

What the Economics Actually Produce: Three Roles That Don't Yet Exist at Scale

Map these mechanisms onto UK service sectors and a set of conditions appears. Models that couldn't previously operate become viable. Those models need people in roles that don't exist at scale yet.

Continuous healthcare monitoring is the clearest case because the price, continuity, access, and personalisation conditions are all moving at the same time, and the human premium requirements are load-bearing across accountability, behaviour change, and relational categories. The roles it generates are entirely new. They aren't substitutions for existing jobs. Three are worth naming because the economics points directly at each of them.

The Continuous Care Navigator is the human layer between patients and the AI monitoring system, holding a caseload of people whose data is being tracked continuously, and surfacing the moments that need human attention. When a pattern changes meaningfully. When a patient needs a call. When a family member needs reassurance. When an escalation has to land somewhere that a clinician can take responsibility for it. The AI ingests data, establishes baselines, detects deviations, ranks cases by urgency, handles documentation. The Continuous Care Navigator does the parts of the job AI can't deliver convincingly: the trust, the translation, the behaviour change conversations, the noticing of things that aren't in the data because they were never captured. At realistic adoption rates, this is a six-figure employment category at UK scale.

The Care Plan Outcome Specialist owns the gap between medical advice and what patients actually do with it. The plan exists. The instructions are clear. The adherence is the problem, and it's almost always a problem because of something practical or emotional the plan didn't account for. Cost. Transport. Fear. Family dynamics. Shame. Lack of confidence about what the advice actually meant. The Care Plan Outcome Specialist is the role that solves those problems in the real world so the clinical plan can actually work. AI tracks the milestones and surfaces non-adherence patterns. The human gets it sorted, person by person, conversation by conversation. Behaviour change is the entire value the role delivers.

The Health Data Operations Specialist holds the institutional accountability for the data that makes the whole system function. Wearables, lab feeds, pharmacy records, patient-reported inputs, insurance data, GP records, all integrating across multiple platforms with permissions, consent, audit, and clinical usability requirements. AI handles a lot of the data work itself. Someone needs to be responsible when it doesn't work, when integration breaks, when devices drift out of calibration, when permissions need to be reset, when the IT and clinical sides need translating to each other. The role is technical, but the value is in the accountability and the translation, not the technical execution.

On reasonable population ratios, that's tens of thousands of roles at UK scale.

Three roles in one sector. The pattern repeats across financial services, education, social care, mental health support, and the professional services tier that AI is bringing into reach for the SME economy. AI makes the service model affordable or operationally viable. Human roles form at the boundaries where accountability, judgement, translation, and behaviour change are required. Those boundaries shift as capability develops. In some of the most important cases, they expand as total volume grows.

The role categories that follow are consequences of new economics rather than a taxonomy of future job titles. Most deployment thinking happening right now hasn't caught up with that shift, which is part of why the roles emerging from new service models are still poorly understood inside most organisations.

Navigators who help people move through systems that have become too complex to face alone. Continuous support workers providing the human layer around AI-monitored processes. AI-augmented practitioners reaching market segments that were previously uneconomical to serve. Data and operations specialists making AI-enabled systems reliable enough to carry institutional weight. QA and compliance people whose skill set barely exists in the UK talent pool and will face increasing pressure as regulated-sector deployment accelerates. Escalation specialists handling cases too complex, too ambiguous, or too emotionally weighted for a system response. That last category grows as a proportion of total caseload because AI takes the predictable cases first, pushing the residual workload up the difficulty curve.

If you're working through where your own business sits in relation to any of this, 360 Strategy provides AI consulting in Scotland focused on commercial decisions rather than positioning exercises.

The Structural Problem the Market Won't Fix

The aggregate picture is a labour market that looks different rather than smaller. That reading doesn't resolve the distributional problem, and conflating the two is probably the most common failure in the current policy debate.

The early UK evidence on distribution is sobering. Research from King's College London analysed millions of UK job postings and LinkedIn profiles between 2021 and 2025. The findings: firms with high AI exposure had cut total employment by 4.5% on average, with the effect concentrated almost entirely in junior positions, which fell by 5.8%. High-paying firms saw employment drop by 9.6%. High-salary occupations saw job postings fall by 34.2% (Klein Teeselink, 2025). McKinsey's UK analysis tracked vacancies falling 43% between May 2022 and May 2025, with disproportionate impact on graduate-level openings (McKinsey, 2025). The aggregate jobs story is positive over the medium term. The transition story for specific groups of people in specific places isn't.

The people most exposed to near-term displacement and the people best positioned to move into the emerging roles are not, in most cases, the same people. Geography compounds the problem in the UK context specifically. Professional and service sector growth concentrates in cities and in organisations that already have infrastructure for continuous professional development. Administrative and mid-skill white-collar roles, which are among the most exposed to AI substitution over the next decade, are distributed across a much wider range of communities. Many of those communities don't have adjacent economies positioned to absorb what's coming.

Markets won't resolve a spatial mismatch of this kind on their own. The UK's institutional response has been inadequate. Deployment is accelerating regardless. The pilot phase is over and the institutional response to workforce impact hasn't kept pace. Retraining at the required scale hasn't happened. The apprenticeship levy has been legitimately criticised for most of its existence. Skills England is the right framing operating on the wrong timescale.

Scotland has at least begun to engage with this seriously. The Scottish Government's AI Strategy 2026-2031 names workforce transition explicitly and identifies the University of Strathclyde's Socially Progressive AI Lab as one of the institutions positioned to support evidence-led policy on AI's impact on the Scottish economy (Scottish Government, 2026). The Fraser of Allander Institute at Strathclyde continues to produce some of the better economic analysis of the Scottish labour market available anywhere. Whether the funding, the urgency, and the implementation catch up with the framing is the open question.

The economic conditions for substantial job creation in an AI-enabled economy are present and legible, if the analysis follows demand-side economics rather than staying on the supply side. Alongside those conditions, the AI strategy Scotland needs is the policy and training infrastructure to convert them into real, accessible employment for the people who need it most. An honest reading of the current institutional landscape doesn't supply a reassuring answer on that front.

The jobs exist in the economics. Getting people into them is a different kind of problem entirely, and it isn't one the AI industry has shown much appetite for owning.

Mark Evans MBA is founder of 360 Strategy, an AI strategy and consulting practice based in Scotland.

References

Available on request.

Frequently Asked Questions

Will AI eliminate jobs in Scotland?

Not on net. The World Economic Forum's 2025 Future of Jobs Report projects 92 million roles displaced globally by 2030 against 170 million created, a net gain of 78 million. The same logic applies in Scotland, though distribution will be uneven. Mid-skill white-collar roles, particularly junior administrative and analytical positions, are the most exposed in the near term. Scottish employment overall is likely to grow as new role categories form around AI-enabled services. The transition will be hard for specific occupations and specific places.

What new jobs will AI create in the UK?

The emerging role families are navigators (helping people move through complex systems), continuous support workers (the human layer around AI monitoring), AI-augmented service operators reaching new market tiers, data and operations specialists, QA and compliance roles in regulated sectors, and escalation specialists handling the cases AI can't manage. In healthcare specifically, three roles stand out: the Continuous Care Navigator, the Care Plan Outcome Specialist, and the Health Data Operations Specialist. Each represents a function the previous labour economics couldn't support at scale.

How will AI affect UK SMEs?

AI is bringing professional services that were previously priced out of reach into the budget range of small businesses for the first time. About 5.5 million UK SMEs make up a market that has been largely dormant for retained marketing, legal, financial, and strategic services. The Affordability Unlock creates demand for AI-augmented service providers and at the same time compresses margins for traditional delivery models. The competitive picture for SMEs depends heavily on how quickly they integrate AI into their own delivery and decision-making.

What is the Affordability Unlock in AI economics?

The Affordability Unlock is the demand expansion that happens when AI reduces the cost of an existing service to a price point where new buyers can finally access it. Professional services dropping from £5,000 to £500 is the clearest example. The market doesn't shrink. It broadens to include buyers who were always present in latent form but priced out. The Affordability Unlock expands existing role categories to serve new market tiers, generating employment in AI-augmented professional services rather than reducing it.

What is the Possibility Unlock in AI economics?

The Possibility Unlock describes service models that couldn't operationally exist before AI made them viable, regardless of price. Continuous preventative healthcare. Always-on financial life support for households who never previously qualified for ongoing advice. Personalised adaptive education at scale. The Possibility Unlock creates economic activity that is genuinely additional rather than redistributing from existing markets, and it generates role categories that don't yet exist in meaningful numbers because the underlying service models haven't been viable until now.

What is the human premium in AI economics?

The human premium is the portion of economic value that stays attached to human involvement even when AI can perform the underlying task. There are seven categories: relational (accumulated trust and context), embodied presence (physical co-location), trust (validating a recommendation before acting), accountability (someone responsible when things go wrong), translation (turning intent into AI-usable instruction), behaviour change (compliance through human relationship), and provenance (the value of human authorship). These are economic mechanisms, not soft preferences, and they create durable demand for human roles in any service market.

Which UK sectors will gain the most jobs from AI?

Healthcare, professional services for the SME economy, education, mental health support, social care, and personal financial guidance all show the conditions for substantial role creation through both the Affordability Unlock and the Possibility Unlock. The new roles cluster at the boundaries where human accountability, judgement, behaviour change, and translation are essential. Regulated sectors will see the strongest growth in QA, safety, and compliance roles where AI system understanding meets regulatory knowledge.

How will AI change healthcare jobs in the UK?

The shift from episodic to continuous care, made possible by AI monitoring, creates three new role categories. The Continuous Care Navigator handles the human layer between patients and AI monitoring systems. The Care Plan Outcome Specialist owns the gap between medical advice and real-world execution. The Health Data Operations Specialist maintains institutional accountability for the data infrastructure that makes continuous care safe and trusted. At UK scale, realistic adoption could mean several hundred thousand new healthcare roles over the next decade, alongside redeployment of existing clinical staff toward higher-judgement work.

Will AGI eliminate the new AI-era jobs too?

The assumption behind this question is that labour demand is purely about task capability. It isn't. The seven human premium categories create durable demand for human delivery that doesn't transfer to AI even when AI can technically perform the underlying task. Accountability requires a human in the regulatory chain. Behaviour change requires a human relationship to produce adherence. Trust, translation, and provenance all attach economic value to human involvement independent of capability. AGI changes the cost of capability. It doesn't dissolve the market mechanisms that attach value to human delivery.

How should Scottish businesses prepare for AI workforce changes?

The starting point is an honest audit of which parts of your service delivery belong to the Affordability Unlock (existing markets you can now reach) and which sit closer to the Possibility Unlock (new service models you couldn't previously offer). Next, identify where the human premium is doing real work in your offering. Those are the roles least at risk and the easiest to defend commercially. The third piece is mapping where the new role categories your sector needs sit relative to your current team, and starting to build those capabilities now. Waiting until the market has organised around them is the expensive option.