UK workforce outlook to April 2028: AI may narrow the entry routes employers need

The thesis: Britain’s workforce problem is shifting from labour scarcity to capability formation

**Forecast horizon: autumn 2026 to April 2028.**

The UK is not moving from labour shortage to labour abundance in any simple sense. It is moving into a more selective hiring market at the same time as AI changes the task mix within administrative, professional, customer-service and knowledge-work roles. That combination creates a risk that is easily missed in headline vacancy data: the erosion of the entry routes through which people acquire occupational judgement, employers develop future supervisors, and local economies turn education into sustained work.

The immediate market context is cooler. UK vacancies were estimated at 702,000 in June to August 2026, down 4.9% on the year and 10.9% below the pre-pandemic level. There were 2.5 unemployed people for every vacancy, while the vacancy rate was 2.1%; the vacancy estimate carries an approximate 95% confidence interval of plus or minus 32,000. (Office for National Statistics, 2026) ↗ Payrolled employment was provisionally 145,000 lower than a year earlier in August 2026, and unemployment was 4.9% in May to July. (Office for National Statistics, 2026) ↗

Yet a looser market is not the same as readily available capability. Skills England reports that 27% of vacancies in the 2024 Employer Skills Survey were hard to fill because of skills shortages. That was below the 36% recorded in 2022, but remained above the pre-pandemic position. Construction, education and manufacturing were notable pressure points, while demand was also high in occupations linked to information and communication, property and finance. (Skills England, 2026) ↗

**Sanctuary interpretation.** The practical challenge for the next 18 months is likely to be making fewer external hires more productive while retaining the capacity to build future talent. Employers may respond to caution by reducing junior intake, leaving posts vacant or automating routine work. Those choices can improve short-term cost control. But, if they are not accompanied by redesigned learning and supervision, they can reduce the supply of people able to take on complex, customer-facing, technical and management work later.

This matters particularly to SMEs, local service businesses and high streets. Smaller firms often provide first jobs and informal occupational learning, but have less management time and less slack to design alternatives when recruitment or training is cut. The policy and business issue is therefore not simply how many jobs AI displaces. It is whether organisations continue to produce the capability on which their own productivity depends.

Why entry routes are exposed—and why the causal story needs care

Three mechanisms make the entry-route issue plausible.

**First, recruitment restraint is concentrated where routes into work are often formed.** Vacancies in businesses with one to nine employees fell by 7,000 quarter on quarter to 91,000 in June to August 2026, and were 23.5% lower than a year earlier. The ONS also reported continuing employer feedback that higher labour costs may be constraining small-business recruitment. (Office for National Statistics, 2026) ↗ A reduction of this scale matters beyond the SME segment: small firms provide local work experience, first managerial responsibility and progression opportunities that are less visible in national occupational statistics.

**Second, AI can compress the lower-risk tasks on which junior learning has traditionally relied.** The Bank of England’s Agents reported productivity gains from AI in software development, finance, administration, customer service, professional services and content creation. They also reported lower demand for some entry-level, administrative and junior roles, including work involving document preparation, invoice processing and basic analysis. However, the report is clear that realised gains vary substantially with implementation, employee capability and oversight. (Bank of England, 2026) ↗

The mechanism is important. Routine drafting, searching, processing and summarising are not merely outputs; they can be the practice ground through which a new recruit learns standards, exceptions, customer context and professional judgement. Removing them does not automatically remove the need for junior workers. It does mean that managers must deliberately create other forms of supervised practice: live casework, quality review, shadowing, escalation handling and progressively delegated decision-making.

**Third, the evidence should not be overstated.** Skills England cites research associating higher AI exposure with falls in junior roles and vacancy posting, as well as evidence of weaker graduate and entry-level online adverts. But it explicitly cautions that post-pandemic adjustment, weak demand and the economic cycle mean the slowdown cannot yet be cleanly attributed to AI. (Skills England, 2026) ↗

That caution is material. Firms may be reducing entry hiring because demand is weak, labour costs are higher, or recruitment has become more risk-averse—not because a model has replaced a job. Conversely, AI may support entry hiring where it enables a small team to win more work, improve service capacity or shorten the time before a recruit can contribute. The central forecast is therefore conditional: AI becomes an entry-route problem when employers remove junior tasks faster than they create credible replacements for learning and progression.

The outlook to April 2028: three scenarios, not a point forecast

**Base case: selective hiring and uneven productivity gains.** This is the most credible scenario. Organisations hold aggregate headcount broadly flat, backfill selectively and prioritise roles that protect revenue, compliance, service quality and scarce technical capacity. The Bank’s Agents reported broadly flat headcount expectations over the following 12 months, targeted recruitment in professional and IT/digital services, and continuing efficiency efforts in response to higher labour costs and weak demand. (Bank of England, 2026) ↗

Under this case, routine junior roles remain under pressure, but demand persists for people who can validate outputs, resolve exceptions, work with customers, operate safely and coordinate across functions. Pay growth moderates at an aggregate level without disappearing: regular earnings grew by 3.5% annually in May to July 2026, while the Bank’s Agents reported a 3.5% weighted average pay settlement for 2026. (Office for National Statistics, 2026) ↗ (Bank of England, 2026) ↗ The planning implication is differentiated labour cost rather than a uniform wage freeze.

**Upside case: demand recovery turns experimentation into augmentation.** If client demand and business confidence improve, firms that have connected AI tools to dependable workflows may use the resulting capacity to grow output rather than simply remove work. Hiring would probably still be targeted rather than broad-based, with stronger demand for technical, customer-facing and operational roles. Entry recruitment could recover where employers can see a credible path from trainee work to productive contribution. The relevant signs would be a stabilisation in small-employer vacancies, stronger recruitment conversion and AI training tied to live workflow deployment.

**Downside case: cost pressure becomes capability erosion.** A further deterioration in demand, renewed cost shocks or persistent uncertainty could extend natural-attrition strategies, reduce discretionary training and increase reliance on contingent capacity. The Bank’s Agents already identified higher labour costs, subdued demand and geopolitical disruption as contributors to hiring caution. (Bank of England, 2026) ↗ In this scenario, junior intake falls faster than work is redesigned, intensifying competition for remaining entry roles and weakening the future supply of experienced staff.

These scenarios are not predictions of occupational collapse. They reflect an implementation lag: accessible AI can be adopted quickly, but robust processes, managerial confidence, governance and development pathways take longer to build. The more quickly employers close that gap, the more likely AI is to augment rather than hollow out workforce capability.

What leaders should measure: from vacancy counts to capability production

Headline vacancies are a useful temperature check, but they do not show whether an organisation is still producing talent. A practical dashboard should connect recruitment, task design, learning and service quality.

1. **Small-employer and entry-level recruitment trends.** Track vacancies, applications, time-to-fill and offer acceptance separately for roles that offer first entry or early progression. The marked decline among the smallest firms is a warning that local entry routes may weaken before the aggregate market does. (Office for National Statistics, 2026) ↗

2. **Entry-route conversion.** Measure applications-to-interview, interview-to-offer, offer acceptance, six-month retention and progression to independently productive work. Intake alone is not a success measure if recruits do not receive enough supervised practice to develop.

3. **Task displacement and task creation.** For each AI-enabled workflow, record which tasks have been removed, which checking, exception-handling and customer tasks have been created, and which career stage performs them. This is more decision-useful than asking whether AI has replaced a role.

4. **Demonstrated, role-based AI capability.** Skills England identifies technical, non-technical and responsible-and-ethical AI foundation skills, including the ability to use automation, configure tools safely and analyse information. (Skills England, 2026) ↗ Organisations should assess competence in the live workflow, not completion of a generic course.

5. **Quality, rework and assurance.** Faster drafting or processing is not a productivity gain if it creates avoidable corrections, escalations, complaints or risk. The Bank’s Agents highlighted concerns about inaccurate outputs and inauthentic applications. (Bank of England, 2026) ↗ Time saved should therefore be reported alongside error, customer and assurance measures.

**Sanctuary recommendation.** Establish a small workforce-redesign group spanning operations, HR, learning, technology and frontline management. Its initial task should be narrow: identify three to five workflows where AI is already changing junior work, then define the replacement learning architecture. That should include supervised live work, structured quality review, peer feedback, exposure to exceptions and clearly staged responsibility. This is less glamorous than a large transformation programme, but it is where productivity and employability become operationally connected.

The strategic choice: use AI to accelerate judgement, not merely remove tasks

The strongest response is neither to preserve every routine task nor to assume that a larger applicant pool has resolved recruitment. It is to use AI to move early-career workers more quickly towards higher-value activity while preserving the practice, feedback and accountability through which competence is built.

The evidence on workforce development supports a practical approach. The *Skills for AI* research argues for development that is practical, accessible, integrated into work, modular and sustainable rather than a one-off exercise in generic training. (Skills for AI, 2026) ↗ That is especially relevant for smaller employers, which may lack the volume of recruits, specialist trainers or protected management time to build a complete pathway alone. Shared provision through employer groups, colleges, providers, sector bodies and anchor institutions can help pool demand for short modules, placements, projects and portable evidence of capability. Skills England similarly identifies SME navigation, collaboration and clearer links between skills investment, retention and business performance as priorities. (Skills England, 2026) ↗

There are clear tests for this analysis. The thesis would weaken if entry-level and small-employer vacancies recover broadly without deterioration in progression, or if employers restore junior recruitment at scale because AI creates more supervised work than it compresses. It would also weaken if widespread workflow deployment produces negligible productivity gains. Leaders should watch those signals rather than treating the forecast as settled fact.

For now, the decision is not simply how many people to hire. It is whether every redesigned role still creates the judgement, technical confidence and management capacity the organisation will need next. Employers that can answer that question well will be better placed to turn AI adoption into durable productivity and stronger local economic value.