by Claude Opus 5.5

What are the main pathways by which AI could change wages and bargaining power across low-, mid-, and high-skill work?

AI affects pay through six main channels. It displaces people from tasks, raises productivity that may or may not be shared, compresses skill differences within jobs, creates scarcity premiums for complementary skills, shifts income from wages to profits, and changes bargaining power through outside options and monitoring. Early evidence points to premiums at the top for scarce AI-related skills and pressure on commodity cognitive and routine office work. The least visible risk is to the pay ladder of early-career and mid-skill workers. In 2026, however, UK wage growth is still driven by the economic cycle, inflation and the minimum wage, not by AI.

Six pathways

1. Task displacement

When AI takes over tasks, the people who did them compete for the remaining work, which pushes down pay for that skill. The cleanest evidence comes from freelance markets, where prices adjust quickly. Hui, Reshef and Zhou found that after ChatGPT’s release, writing freelancers on a large online platform saw monthly earnings fall 5.2%. Image freelancers saw a 9.4% fall after image generators arrived. Highly rated freelancers were not shielded. Salaried pay adjusts more slowly. It shows up first as fewer vacancies and weaker starting salaries rather than pay cuts.

2. Productivity sharing

AI can make each worker more productive, but productivity raises wages only if workers can capture part of the gain. Capture depends on outside options, competition between employers and collective bargaining. In the UK these are thinner than many assume. Trade union density among employees was 22.4% in 2025, but only 12.1% in the private sector against 48.5% in the public sector. For most private-sector workers, a productivity gain becomes higher pay only if rival employers bid for their skills.

3. Skill compression within jobs

Several studies find AI helps weaker or newer workers most. Brynjolfsson, Li and Raymond found a 34% productivity gain for novice customer-support agents against minimal gains for the most experienced. Dell’Acqua and colleagues found below-average consultants improved by 43% and above-average ones by 17%. This narrows the gap between workers, which could compress pay within occupations and reduce the premium for experience. It could also lower barriers to entry, which helps outsiders but erodes incumbents’ premiums.

Autor and Thompson (2025) add a crucial distinction. When automation removes the less expert tasks in an occupation, the remaining work becomes more specialised, so wages tend to rise and employment falls. When it removes the expert tasks, the job becomes easier to do, so wages fall and employment can rise. The same technology can raise pay in one occupation and cut it in another.

4. Scarcity premiums for complementary skills

Where AI raises the value of a skill that is in short supply, pay rises. PwC’s 2026 AI Jobs Barometer puts the UK wage premium for jobs requiring AI skills at 34.2% in 2025, up from 11% in 2024. It ranges from 12% in government to 64% in consumer markets. UK postings for specialist AI roles rose 61% to 180,000. IT shows the same pattern. Indeed reports posted wage growth of 7.2% for IT systems roles against 3.9% across all postings, and Adzuna reported average advertised IT salaries up 16.8% on the year to £67,771 in July 2026 (as reported).

Read these premiums with care. They compare job adverts that ask for AI skills with adverts that don’t. Those adverts may differ in seniority, sector and location, and in a weak market a premium can coexist with fewer jobs. The 2024-to-2025 jump looks large enough that composition effects are likely part of the story.

5. The labour share

If AI substitutes for labour across many tasks, more of national income goes to the owners of capital, models and data, even if total income rises. The OBR’s March 2026 structural unemployment scenario models this explicitly. Technology substitutes for labour, equilibrium unemployment rises to 5.5%, “real earnings do not rise with productivity” and the labour share falls. Its July 2026 Fiscal Risks and Sustainability report warned that AI could shift GDP “from (more highly taxed) labour to (lower-taxed) profits”. The historical precedent is Robert Allen’s “Engels’ pause”. In the first half of the 19th century, British output per worker grew while real wages stagnated and the profit share rose. It took decades before wages caught up.

6. Bargaining power, outside options and monitoring

AI changes bargaining power even where it doesn’t change tasks. If the threat of replacement becomes credible, workers’ leverage falls before any job goes. AI-driven performance monitoring and algorithmic management make effort more observable and standardise work, which weakens informal discretion. The Employment Rights Act 2025 contains no AI- or surveillance-specific provisions. The main UK constraints are data protection law, including the Data (Use and Access) Act’s safeguards on automated decisions in force since February 2026 and the ICO’s 2023 monitoring guidance (now under review), together with equality law. The TUC has called for mandatory AI impact assessments and human review of high-risk decisions. On the other side, AI can strengthen workers who can sell their output directly or switch employers more easily.

How the pathways land by skill level

  • Main exposure. Low-skill: Scheduling, monitoring, some customer contact. Mid-skill: Admin, customer service, junior professional and back-office work. High-skill: Drafting, research and analysis sub-tasks.

  • Main wage pressures. Low-skill: Algorithmic management; automation made more attractive by rising labour costs. Mid-skill: Displacement, slower hiring, compression of experience premiums. High-skill: Commoditisation of routine expertise.

  • Main supports. Low-skill: National Living Wage floor; physical and care work hard to automate. Mid-skill: Demand expansion where services get cheaper; new oversight roles. High-skill: Scarcity premiums; accountability; client relationships.

  • Bargaining. Low-skill: Weak (low private-sector density). Mid-skill: Weakening if tasks become modular and measurable. High-skill: Strong for those with portable reputations.

For low-paid work, the National Living Wage matters more for pay than AI does. The April 2026 rise was 4.1%. The Bank of England’s Agents report that rising NLW and employer National Insurance costs have contributed to headcount reductions in exposed consumer sectors. That raises the incentive to automate at the margin even where AI is not yet the cheaper option.

For mid-skill work, the main risk is not falling pay for people in post but fewer openings and weaker progression. Morgan Stanley’s survey found UK cuts concentrated in roles requiring two to five years’ experience (as reported). That is exactly the rung where pay used to rise fastest.

For high-skill work, the gains flow to those who combine domain expertise, accountability and the ability to direct AI. Professionals whose value lay in producing standard analysis may find that part of their work is priced like a commodity.

The 2026 UK wage picture in context

Regular pay grew 3.5% in nominal terms in May–July 2026, or 0.6% after CPIH inflation. That is down from 4.5% in the January release. Private-sector regular pay grew 2.9%. Nothing in these aggregates points to an AI effect. They are explained by a softer labour market, unemployment at 4.9% and 702,000 vacancies. The Bank of England’s September 2026 minutes mention AI only in relation to technology goods demand and prices, not the labour market. Any AI wage effects will appear first in the composition of pay: premiums for some skills, flatter starting salaries in exposed graduate routes, and wider dispersion within occupations.

What to watch

  • Starting salaries and the number of openings in exposed graduate schemes, such as accountancy, law and software.

  • Pay dispersion within occupations, not just averages between them.

  • Whether AI-skill premiums persist once adverts are compared like for like.

  • Labour-share data as productivity gains, if they come, accumulate.

Sources

From AI and Jobs: UK, October 2026