UK workforce outlook for 2027–28: selective hiring, AI and the entry-route test

Thesis: a looser labour market can still be difficult to staff well

**Forecast horizon: April 2027 to April 2028.** The UK labour-market question is changing. The central issue is increasingly unlikely to be whether employers can attract any applicants at all; it is whether they can organise people, technology and management attention to deliver work reliably at a sustainable cost.

The evidence supports a selective-hiring outlook. UK vacancies were estimated at 702,000 in June to August 2026, 36,000 lower than a year earlier and the lowest level since February to April 2021. Vacancies among businesses with one to nine employees fell by 23.5% over the year, to 91,000. Business feedback collected by the Office for National Statistics also indicated that higher labour costs were constraining recruitment. This points to tighter hiring approvals—particularly among small employers—rather than a straightforward disappearance of labour demand (Office for National Statistics, 2026) ↗.

Yet a softer vacancy market does not automatically mean that work has become easy to staff. The 2024 Employer Skills Survey recorded 250,500 skill-shortage vacancies, equivalent to 27% of vacancies. Technical and practical skills contributed to 87% of those shortages. That rate was well below the exceptional conditions of 2022, but remained above the range seen before the pandemic. Employers may therefore receive more applications for some generalist jobs while continuing to struggle to find people who can perform specific work to the required technical, commercial or regulatory standard (GOV.UK, 2025) ↗.

**Sanctuary assessment.** The meaningful divide in 2027–28 is likely to be between employers that treat recruitment as a headcount transaction and those that treat it as part of a wider operating model. The latter will define which tasks require human judgement, which can be augmented, where accountability sits and how new entrants acquire competence. This matters especially for SMEs, high-street firms and local service businesses, where weak induction, rework and informal technology use can consume scarce management time quickly.

What the evidence says—and the limits of the forecast

Four evidence streams point towards caution rather than a simple employer-friendly market.

First, recruitment demand is weakening and smaller businesses appear especially exposed to cost pressure. Vacancy estimates carry sampling uncertainty, so a single monthly movement should not be overinterpreted. However, the year-on-year fall in vacancies, alongside the sharper reduction among microbusinesses, is a meaningful directional signal. It may reduce competition for applicants in some local labour markets, but it can also reduce the first-job, part-time and career-change opportunities that small firms often provide (Office for National Statistics, 2026) ↗.

Second, cost pressure is moderating rather than vanishing. The Bank of England’s September 2026 Agents’ summary described employment intentions as broadly flat and recruitment difficulties as below normal. Cumulative 2026 pay settlements averaged 3.6%, and contacts expected 2027 settlements to be similar or slightly lower. The same report, however, identified continuing pressures from inflation, union activity and recruitment difficulty in particular occupations (Bank of England, 2026) ↗. This is business intelligence, not a universal pay forecast: statutory wage changes, sector agreements, local labour markets and occupational shortages will produce different outcomes.

Third, AI use is widespread enough to affect work design, but not sufficiently mature to justify sweeping claims about displacement. Around 35% of businesses with 10 or more employees reported using at least one AI technology in June 2026, up from around 12% in late 2023. Among adopters, however, the average number of technologies used rose only modestly, from around 1.4 to 1.6. Adoption also varied sharply by sector, while employee-reported use exceeded business-reported use. That gap is consistent with some activity being individual, experimental or weakly governed rather than embedded in redesigned operating processes (Office for National Statistics, 2026) ↗.

Fourth, employers have less room for unfocused training responses. Real-terms employer training expenditure was £53.0 billion in 2024, down from £59.0 billion in 2022 and 18.5% below its 2011 level. Only 59% of establishments had trained staff in the previous 12 months. AI skills made up 5% of reported digital skills gaps, with notable differences by sector and employer size (GOV.UK, 2025) ↗.

Taken together, this evidence does not support a forecast of mass AI-driven job loss, nor does it establish that skills shortages will remain at precisely their current level. It supports a narrower conclusion: employers are seeking tighter control of labour costs while many still lack the training investment, process discipline and management capacity required to turn technology adoption into dependable productivity.

The mechanism: task redesign raises the value of judgement and supervision

The most plausible near-term effect of AI is task recomposition rather than wholesale role elimination. The ONS found that most AI-adopting businesses reported no change in overall workforce headcount. It also cautioned that business-level measures do not consistently distinguish light user-level activity from AI embedded deeply in production processes. Adoption data should therefore not be treated as evidence that jobs have already been displaced (Office for National Statistics, 2026) ↗.

The analytical point is not that every role will be transformed. It is that where employers use AI to accelerate repeatable activities—such as drafting, information retrieval, scheduling, basic communications or elements of analysis—the residual work can become more demanding. Someone still needs to frame the task, verify the output, understand the customer or case context, protect information, handle exceptions, escalate risk and take responsibility for the decision.

This creates a management trade-off. Removing low-value repetitive work can increase capacity and improve service. But removing junior work indiscriminately can also remove the practical tasks through which people learn standards, judgement and organisational context. Employers may then recruit more selectively for experienced workers while creating too few pathways for people to become experienced. The result is not necessarily fewer jobs overall; it is a weaker internal supply of future supervisors, specialists and managers.

Management capability determines whether this trade-off is handled well. ONS research found that firms with stronger management practices were more likely to follow through on planned AI adoption. Technology adopters were also associated with higher turnover per worker, although the research does not show that technology itself caused the difference (Office for National Statistics, 2025) ↗. The practical implication is not that every business should automate more quickly. It is that effective adoption requires managers who can select worthwhile use cases, set acceptable practice, test quality, allocate accountability and coach people through changed work.

For many small firms, this is the binding constraint. A software subscription may be affordable; documenting a process, checking outputs and developing a new employee require time that owner-managers often lack. Local productivity policy and business support should therefore treat management practice, workforce development and technology adoption as connected issues rather than separate programmes.

Three scenarios for 2027–28

**Base case: selective hiring and limited workflow integration.** Replacement hiring outweighs broad expansion, while employers continue to control labour costs closely. AI shifts gradually from personal productivity use towards a limited number of governed workflows where service quality, speed or risk can be measured. Demand is strongest for people who combine occupational knowledge, communication, data handling and judgement with AI-supported execution. Broadly flat employment intentions, below-normal recruitment difficulty and rising but still shallow AI adoption make this the most evidence-consistent scenario (Bank of England, 2026; Office for National Statistics, 2026) ↗ ↗.

**Upside case: productivity releases capability investment.** Better process design enables firms to reinvest time savings in service, innovation and growth. Hiring recovers first in technical, commercial, operational-improvement and supervisory roles. This would require more than greater use of AI tools: it would require evidence of sustained demand, meaningful workflow integration and investment in training and management.

**Downside case: defensive cost control and a thinner entry pipeline.** Weak demand persists, particularly among smaller firms. Employers defer hiring and deploy technology principally to constrain costs, without enough process ownership or quality assurance. Entry-level and part-time opportunities narrow even as experienced, customer-critical and technical workers remain difficult to replace. The Bank’s Agents reported concerns in some local labour markets about fewer entry-level and part-time opportunities for young people. This is qualitative intelligence rather than a national causal estimate, but it is a relevant warning for employers, colleges and employability providers (Bank of England, 2026) ↗.

The scenarios are deliberately conditional. Vacancy data have sampling uncertainty; the Agents’ report is informed intelligence rather than a representative survey; and current AI measures reveal more about whether businesses use AI than about the depth, quality or consequences of that use. The common strategic choices are nevertheless robust across all three scenarios: clarify human accountability, build management capability and preserve progression into higher-value work.

A practical agenda: hire for capability, not proxies

**Sanctuary recommendations**

**1. Review the work before approving the role.** For each proposed vacancy, identify tasks that require human judgement, tasks that can be augmented and tasks that should be removed or simplified. The business case should specify the intended outcome—revenue, service quality, compliance, risk reduction, capacity or capability transfer—not merely the number of hours required. This prevents recruitment from becoming a substitute for process design.

**2. Assess demonstrated capability rather than inflated proxies.** Structured interviews, relevant problem-solving exercises and proportionate paid work samples offer better evidence than unnecessary degree requirements, arbitrary years-of-experience thresholds or long software lists. This does not mean lowering standards. It means testing the capabilities the role genuinely requires, which can widen access to people with transferable experience.

**3. Make the first six months a managed capability process.** New recruits need operating standards, purposeful check-ins, feedback on output quality, explicit AI-use boundaries and defined learning tasks. A 30-, 90- and 180-day review structure makes probation an active investment in performance, retention and risk management rather than passive observation.

**4. Focus training on high-value workflows.** Training should connect technical task knowledge with checking, escalation, customer impact, information handling and responsible technology use. Generic prompt-writing sessions will rarely change performance on their own. With real-terms training expenditure declining, prioritisation is essential; indiscriminate cuts, however, risk worsening the skills constraints employers already report (GOV.UK, 2025) ↗.

**5. Protect entry routes deliberately.** Where routine junior work is reduced, replace it with supervised exposure to quality assurance, customer cases, data checking, exception handling and workflow improvement. Employers, colleges and employability providers should align placements and first roles with live business processes, not obsolete task lists. This is both an inclusion measure and a resilience measure: it creates a future pool of people who understand how work is actually delivered.

The strategic opportunity is not simply to employ fewer people. It is to create workforce systems that can use technology without hollowing out judgement, supervision or progression. Employers that connect role design, management practice and entry routes will be better positioned whether the economy strengthens, stagnates or weakens further.