Artificial intelligence, business and the future of the workforce
The Trades Union Congress (TUC) brings together the 5.3 million working people who make up our 47 member unions. We support unions to grow and thrive, and we stand up for everyone who works for a living.
The rapid adoption of AI in the workplace sits at the heart of our core mission to protect and advance the interests of working people. As AI reshapes tasks, job roles, pay structures and working conditions across virtually every sector of the economy, the TUC has both the representational mandate and the practical expertise – drawn from our member unions operating across diverse industries – to speak to the real-world impacts these technologies are having on workers.
Our submission to this consultation argues that while AI adoption is widespread, it remains shallow and has yet to deliver the promised productivity gains, largely because workers are routinely excluded from decisions about how these technologies are introduced.
We highlight a range of uneven and often harmful impacts on workers, including job displacement concentrated among junior roles, wage compression in some sectors, work intensification, skill erosion through cognitive offloading, and the risks of opaque algorithmic management to fairness, privacy and employment rights. Creative workers and gig economy workers are identified as particularly vulnerable.
To address these challenges, we have set out a five-point pro-worker AI innovation strategy, calling for meaningful worker and union involvement in AI governance, stronger collective bargaining rights over technology adoption, updated regulatory frameworks to protect workers from algorithmic harms, fairer distribution of productivity gains, and robust support for workers navigating job transitions.
Call for evidence: Artificial Intelligence, business and the future of the workforce
1. Understanding AI adoption and effectiveness
Generative AI tools have seen a rapid and sizable uptake in usage. However mere usage of AI tools may be of limited significance when considering AI’s impact on work and the wider economy. A study by BCG found that while ‘most organizations are preoccupied with inputs, such as the number of logins to AI tools or the amount of time spent using them,’ the rate of usage had limited bearing on improving quality or productivity of work, and in cases led to a decline. 1 While certain sectors, such as software development have seen deep adoption, on the whole AI adoption can be described as broad but shallow.
Consequently, promised productivity gains (and for the AI developers, profitability) have yet to materialise. This is prompting a greater recognition that adoption necessitates moving beyond top-down target setting and towards deeper consideration of what technologies are being used, by whom, for what purpose, and under what conditions.
AI adoption can only be effective with due consideration for workforce development, workflow reorganisation and related organisational practises. This requires effective worker voice, participation and negotiation to leverage the expertise, practical experience and buy-in of the workforce expected to use these tools.
For policymakers, employers, and workers, an appreciation of context is also key. AI is not a single technology or use case – with each context raising a variety of technological, ethical and operational concerns. The use of edtech by teachers is distinct to the use of AI video generation tools in marketing or the use of a fined tuned Large Language Model (LLM) in legal settings.
While AI is often treated as a ‘magic bullet’ with board and shareholders pushing for efficiency and cost savings, AI technologies require careful integration, and have implications for tasks, skills, job quality and design.
A key barrier to meaningful adoption is the routine exclusion of workers from decisions about technology that reshapes their work. 2 Worker involvement is often treated in public debate as a brake on innovation. However, workers and their unions have direct understanding of the practical realities of tasks, workflows and risks in a way boards and third-party tech companies do not. This makes workers well positioned to determine whether and where new technology will be genuinely beneficial, flag reliability issues and errors. and identify needs for skills, training and organisational innovation.
Along with the TUC and unions, the necessity of meaningful worker voice is increasingly recognised by others. Microsoft’s Future of Work 2025 report found that ‘centring worker voice in AI design boosts productivity, satisfaction, and skill growth – driving both business success and worker flourishing.’ 3
To move towards AI adoption that is purposeful, fair and effective, changes will be needed to the current legal regimes and practises of industrial relations and worker voice in the UK.
2. Impacts on work & workers
The impact of AI on work and workers is characterised by general uncertainty, though some specific issues are emerging clearly.
We recognise that AI technologies can have positive applications for workers including health and safety, automation of certain tasks, enabling access to skills and information – all of which can contribute to productivity gains. However, for the purposes of this submission we have mainly focused on challenges and unequal impacts, as they most demand scrutiny.
Below we set out some key themes for consideration.
‘AI exposure’, task substitution and displacement
The most common means to determine the potential impact of AI automation on jobs is the ‘task-based framework’ that analyses the ‘bundle of tasks’ that make up a job and assesses to what extent AI can automate in whole or part those tasks. This has led to many assessments that the UK’s service sector dominated economy is highly vulnerable to substantial displacement. While the task-based analysis is still useful, more recent analysis has complicated this view, positing that AI ‘exposure’ alone does not predict job loss. Instead, displacement depends on three factors:
(1) task complementarity: automating a complementary task can increase a worker’s value;
(2) demand elasticity: whether productivity gains lead to more output and increased labour demand, or to fewer workers producing the same or more; and
(3) job dimensionality: jobs with few core tasks (e.g., truckers, warehouse workers) are far more automation‑sensitive than high‑dimensional jobs, like those in professional service roles where relationship management, judgement and strategy are less able to be automated. 4
Labour market economists have long demonstrated that the amount of work in the economy is not fixed, and that labour market adjustments can create more jobs. Yet this ‘lump of labour fallacy’ is still applied to debates about technological innovation. We agree that AI may produce new types of jobs and demand for labour, as previous rounds of technological change have. But labour market disruption always brings winners and losers, and intervention is necessary to support those whose jobs change and whose skills need to be updated.
Key questions for this wave of change are whether and when new jobs will materialise, what quality they will be, who will be able to access them, and how current and future workers will be supported through potential economic and social disruption.
Wage uplifts in some, wage compression in others
AI’s impacts and capabilities are ‘jagged’ – systems are highly capable in some areas and not in others or are deeply deployed in some areas and not at all in others. Below we set out examples that illustrate the variable impact of AI technologies.
Usage of AI in some sectors (e.g. software development) can augment or complement existing worker and lift workers’ wages by depressing demand for new workers. A 2025 study from Kings College London found that firms whose workforces are highly exposed to AI capabilities reduced their total employment by 4.5% on average, with the effect concentrated almost entirely in junior positions, which fell by 5.8%. Meanwhile, average pay within those companies increased by more than £1,300. 5
In other cases, there can be wage compression. Cities in the USA where self-driving vehicles have experienced significant adoption have seen significant wage compression of drivers remaining in the sector compared to cities without them. 6 Market operators’ strategies and pricing decisions largely determine outcomes in this context – which is pertinent to governance of London’s AV pilots and eventual roll outs.
Workers in the creative industries, especially freelancers, are also experiencing significant impacts. A report surveying 10,000 creative workers recently found that a of third (32%) of illustrators lost commissions or cancelled projects due to GenAI, 86% of authors say GenAI has already reduced their earnings. Among musicians, 73% say unregulated GenAI now threatens their ability to earn a living. 7
Work intensification
New research from University of California found that rather than reducing work, generative AI can intensified it. 8 Employees at a U.S. tech company voluntarily adopted AI tools with the hope of freeing time for higher‑value tasks, but instead found themselves working at a faster pace, taking on a broader set of responsibilities, and extending work into more hours of the day. AI made ‘doing more’ feel easy and rewarding, prompting workers to absorb tasks from other roles and juggle parallel streams of activity, raising expectations for speed and widening the scope of what counted as ‘their job.’ Despite feeling more productive, participants consistently reported feeling just as busy, or even busier, than before.
While the study illustrates productivity gains from AI adoption, it also evidences the importance of active job redesign to ensure that workers share in these benefits. The self‑reinforcing cycle of acceleration brought significant downsides: workload creep, cognitive fatigue, weakened decision‑making, and early signs of burnout as the initial excitement faded. Organisations, meanwhile, risk misinterpreting intensity as sustainable productivity.
Productivity, cognitive offloading and skill formation
AI automates tasks that workers might otherwise do. While this may be welcome, existing evidence adds nuance. For example, several recent studies have found that generative AI does not reliably increase productivity: one from leading AI benchmarking organisation METR found that experienced software developers took 19% longer to complete tasks using AI than without it. 9
Regarding skill formation and ‘cognitive offloading’, a study by Anthropic found that, coders who used AI assistance had no increase in productivity but a 17% decrease in skill mastery. 10 For many workers, who sell time and skills in return for wages, the prospect of de-skilling is a significant concern. However, some academics have also argued that AI enables a levelling of access to skills and therefore some widening of opportunities.
These studies also contribute to the understanding of what is colloquially called ‘workslop’ – where tasks are rapidly completed using AI but to a poor standard, often needing rectification.
Professional standards, liability and ethics
AI automation can contribute to workload alleviation but is fraught with governance issues, especially in high-risk settings. For example, AI transcription tools are being rapidly adopted across social care, with social workers widely reporting that these tools reduce paperwork and deliver meaningful time‑saving benefits, helping shift their attention back toward direct, relational practice rather than documentation.
However, this research by the Ada Lovelace Institute found that the tools introduce significant new pressures and risks for workers. 11 Social workers currently remain fully responsible and legally accountable for the accuracy of AI‑generated notes, summaries, and case records, creating added cognitive load and professional liability in detecting errors. Workers report concerns about biased outputs and hallucinations, including incorrect or harmful information being introduced into case records, which requires additional oversight and can undermine trust in the tools. Perceptions of reliability and appropriate use vary widely, and there is currently no consensus on when AI transcription is suitable, leaving workers to navigate ambiguous ethical boundaries and uneven guidance.
Autonomy, privacy, fairness, discrimination and employment rights
Algorithmic management, or the use of AI systems to oversee and direct workers, forms a major part of how AI is now used in the workplace and raises several serious concerns for workers. These are set out below.
Workplace discrimination and fairness: AI tools used in recruitment, performance assessment, and dismissal can entrench and even amplify bias, especially when algorithms operate opaquely. This opacity can make discriminatory decisions harder to detect or contest, undermining workers’ fundamental employment rights and protections against unfair treatment.
Health, safety, and wellbeing: Expanding monitoring and data-driven oversight can intensify work, reduce autonomy, and increase stress, contributing to burnout. The combination of constant surveillance and accelerated work processes creates clear risks to both mental and physical wellbeing – risks that current health and safety frameworks are not yet fully equipped to address.
Trade union rights and collective voice: AI can also pose challenges to core trade union freedoms. Data-driven systems may be used to identify or target workers likely to organise, raising concerns about victimisation. A lack of transparency in AI systems can hinder unions’ ability to represent members, engage in collective bargaining, and effectively negotiate with employers. Furthermore, AI-based monitoring and content moderation tools may restrict workers’ communication and organising efforts, potentially interfering with rights protected under Article 11 of the European Convention on Human Rights.
‘Dynamic’ algorithmic pay: Oxford University research found that Uber’s shift to an algorithmic, dynamic pricing system has increased fares while reducing driver earnings. 12 After the algorithm was introduced in 2023, passengers began paying more per trip, yet drivers’ hourly income fell from over £22 to just over £19 before costs, with unpaid time spent waiting for rides increasing. Uber’s commission simultaneously rose from roughly 25% to 29%, sometimes exceeding 50%, with the company taking a larger cut from higher value‑ fares. The study raises serious concerns about transparency, fairness, and the power imbalance created by opaque algorithmic pricing in the gig economy. Where employers refuse to accept that workers are employees, relying on legal loopholes that the TUC wants to see closed, platform economy workers have little recourse to redress or collective negotiation outside of limited provisions within data protection law.
Distribution of income between capital and labour
Growing inequality and fiscal pressure may worsen if more income continues to flow to capital owners rather than workers. A fair and effective tax regime and extension of collective bargaining is needed so that workers can maintain a share of income.
A March 2026 Centre for Economic Policy Research (CEPR) study found that areas across Europe with higher AI-‑related patenting saw a clear decline in labour’s share of income, especially in industrial regions, suggesting that AI is a capital-‑biased technology whose productivity gains mainly benefit capital owners. 13 Similarly, a 2024 analysis from the Philadelphia Federal Reserve warned that generative AI could reduce labour’s national income share by directly replacing human work, posing a structural risk not seen in past technological waves where productivity gains typically boosted workers’ incomes. 14 The UK’s OBR likewise noted in March 2026 that advanced technologies could lower labour’s income share and reduce the tax base, raising concerns about long-term‑ economic stability. 15
3. A pro-worker approach for AI policy, regulation & rights
Government has a crucial opportunity to shape the trajectory of these technologies and their socio-economic impacts through proactive policymaking. The TUC welcomes many of the initiatives of the government to date, including the Responsible AI Adoption Panel and the Future of Work Unit; efforts to support UK based technology providers with strategic use of public procurement; and the ICO guidance on Automated Decision Making and DFE standards for education technology. We also look forward to the government’s Make Work Pay commitment to consult on surveillance technologies.
But a more comprehensive national plan is necessary to meet the scale and complexity of the challenge. Towards that end in August 2025 the TUC set out a five-point paper ‘Building a pro-worker AI innovation strategy’. 16 This is summarised below.
1. Innovation for Good Work
The UK’s world-class research capabilities, funded by billions in public money through agencies like UKRI, currently lack systematic participation from workers. Involving workers and unions in the innovation ecosystem will help steer research toward worker-complementary technologies and productivity-enhancing applications rather than low-productivity automation designed solely to cut jobs.
Policy Recommendations:
- DSIT should mandate worker engagement in the governance of key research institutions including UKRI, Innovate UK, and the Catapult Network.
- DSIT, working with other relevant public bodies, should support publicly funded collaborative research programmes between unions, universities and industry in key sectors (e.g. finance, law, creative industries) to develop worker-centric AI.
- The British Business Bank and National Wealth Fund should be directed to prioritise investments in companies that create good jobs, using equity stakes and employment charters.
- Government should allocate access to the UK’s expanding public AI compute infrastructure toward national priorities that support good work.
2. Returning Public Value to Public Services
Public services are under pressure to adopt AI to reduce costs, but procurement that favours big tech comes with big risks, including ‘vendor lock-in’ and the operational risks of relying on a few dominant global tech firms. Public sector workers are key assets in mitigating these risks, as their expertise is necessary to design effective solutions and avoid algorithmically amplified failures.
Policy Recommendations:
- Government should prioritise in-house development of AI systems, using procurement only where capacity gaps exist.
- Public service providers should coordinate procurement to leverage collective purchasing power and achieve economies of scale.
- Government should require ‘good work’ standards in all public contracts, including union access and recognition.
- Government procurement frameworks should embed worker voice throughout digital transformation – from strategic planning through to deployment – and through a reinstated Public Services Forum.
- Government should expand support for UK AI firms that actively engage with trade unions and create skilled jobs.
3. Securing a Digital Dividend and Productive Investment
AI has the potential for significant productivity gains, but workers must capture a fair share of this ‘digital dividend’. Collective bargaining and corporate governance reform are needed to shift the focus from short-term shareholder profits to long-term productive investment in the workforce.
Policy Recommendations:
- Government has strengthened collective bargaining rights through the Employment Rights Act. Now, it should extend statutory bargaining to cover the introduction of new technologies and AI.
- Sectoral bargaining frameworks should be explored for the application of AI with the goal for promoting high standards of job quality and employment.
- Government should reform corporate governance by amending directors’ duties to prioritise long-term success and requiring worker-elected directors to hold one-third of board seats in firms with more than 250 employees.
- HMT should capture productivity gains for public use through wealth taxation, equalisation of capital gains and income tax, review incentives for investment in labour-augmenting technology, and increase the bank surcharge.
4. Pro-Worker AI Regulation and Upgraded Rights
Urgent action is needed to update laws protecting workers from the harms of AI-powered decision-making, such as intrusive monitoring and biased algorithmic management. A pro-worker approach should also give workers' welfare needs equal standing with consumer welfare in regulatory frameworks.
Policy Recommendations:
- Government should legislate to implement the provisions set out in the TUC AI and Employment Bill, including Workplace AI Impact Assessments, human oversight for high-risk decisions, and strengthened worker data rights.
- Government and regulators must protect the intellectual property and copyright of creative workers from unauthorised AI training. The CMA should challenge dominant firms that impose terms enabling the exploitation of creative materials; the ICO should be given an expanded remit to support collective data rights, enabling unions to help workers exercise control over the data and value they generate.
- Government should improve resourcing of enforcement bodies – including the EHRC and HSE – and provide the CMA and ICO with an explicit pro-worker mandate.
- Government should introduce mandatory human rights and environmental due diligence requirements covering global AI value chains and the environmental impacts of data centres.
- Government should support the development of an independent AI assurance market for workplace technologies.
5. Supporting Workers During Change
While a ‘rapid collapse’ in employment is unlikely, AI will disrupt tasks and sectors unevenly, disproportionately affecting young, female, and regional workers. To avoid the mistakes of past deindustrialization, the state must focus on creating good jobs and supporting workers through transition.
Policy Recommendations:
- DfE, DWP and DSIT should ensure formal trade union representation in the governance, design and deployment of AI skills initiatives.
- Government should require greater employer investment in workforce training to reverse the long-term decline in UK training spending.
- Government should develop an AI workforce strategy for each sector covered by the industrial strategy.
- Government should renew the Union Learning Fund model and expand the role of Union Learning Representatives in delivering AI-related skills.
- Government, in the longer term, should strengthen social security and contributory benefits to provide adequate support during job transitions.
- 1 BCG (2025). ‘The AI Adoption Puzzle: Why Usage Is Up But Impact Is Not’
- 2 Institute for the Future of Work (2024). ‘Firm-level adoption of AI and automation technologies: Case Studies Report’; Microsoft (2025) New Future of Work Report. ’Adoption and Usage’ p.7-16 ; Stanford University Human-Centred Artificial Intelligence (2025). ‘What Workers Really Want from Artificial Intelligence.’
- 3 Microsoft (2025). ‘New Future of Work Report’
- 4 Imas, Alex. Shukla, Soumitra (2026). ‘How Will AI-driven Automation Actually Affect Jobs?’
- 5 Klein Teeselink, Bouke (2005). ‘Generative AI and Labor Market Outcomes: Evidence from the United Kingdom’
- 6 Allon, Gadd (2025). ‘The Impact of Robotaxis on Gig Worker Wages: Early Evidence and Future Outlook’
- 7 Drury, Rachael (2026). ‘Brave New World: Justice for creators in the age of Gen AI’
- 8 Ranganathan, Aruna. Ye, Xingqi Maggie (2026). ‘AI Doesn’t Reduce Work—It Intensifies It’
- 9 METR (2025). ‘Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity’
- 10 Anthropic (2026). ‘How AI Impacts Skill Formation’
- 11 Ada Lovelace Institute (2026). ‘Scribe and prejudice? Exploring the use of AI transcription tools in social care’
- 12 Binns, Reuben et al (2025). ‘Not Even Nice Work If You Can Get It; A Longitudinal Study of Uber's Algorithmic Pay and Pricing’
- 13 CEPR (2026). ‘AI and the distribution of income between capital and labour’
- 14 Federal Reserve Bank Philadelphia (2024). ‘Generative AI: A Turning Point for Labor's Share?’
- 15 Office for Budget Responsibility (2026). Economic and fiscal outlook – March 2026
- 16 Trades Union Congress (2025). ‘Building a pro-worker AI innovation strategy’
Stay Updated
Want to hear about our latest news and blogs?
Sign up now to get it straight to your inbox
