Written submission from Dr Daniel Wheatley, Dr Harriet Clarke, Dr Christian Darko, Dr John Gibney, Dr Benjamin Hopkins, Dr Aikaterini Tavoulari (AIB0040)

Business and Trade Select Committee on Artificial Intelligence, Business and the Future of the Workforce

AI-EMPOWERED Evidence Submission

Dr Daniel Wheatley*, Dr Harriet Clarke*, Dr Christian Darko*, Dr John Gibney*, Dr Benjamin Hopkins* and Dr Aikaterini Tavoulari**

*University of Birmingham **University of Bath

1               Introduction

1.1               The AI-EMPOWERED project, led by researchers from the Work Inclusivity Research Centre (WIRC), University of Birmingham, assesses the implications of artificial intelligence (AI) and automation technologies for the employment opportunities and working experiences of disabled people in the UK.

1.2               Evidence is provided in response to questions under theme 2 (AI adoption), 4 (impacts on work and workers), 5 (skills, education and transitions) and 6 (Government strategy, regulation and rights).

1.3               We will publish a report presenting our full evidence in Summer 2026.

1.4               We propose the following recommendations to UK Government:

Inclusive Governance and Representation

• Ensure disabled people have a central and continuous role in Government enquiries, decision‑making processes, and policy interventions on AI, working through disabled people’s organisations, charities, and representative bodies.

Employer Support and Responsible Adoption

• Develop support and training package for businesses, especially SMEs, to provide minimum level awareness and leadership skills in AI-integrated support for disabled workers, and support in sourcing secure AI solutions and assistive technologies.
• Review reasonable adjustments policy and guidance to respond to changes in workplace digital technologies including how AI‑integrated assistive tools are assessed, implemented, and maintained.
• Embed inclusive design principles for AI and automation technologies requiring sustained commitment from technology developers to accessibility in AI design through inclusive user-led development processes.

Regulation, Rights and Data Governance

• Fund controlled trials of AI-integrated recruitment systems to measure impacts of automated screening, shortlisting and AI-performed interviews on recruitment outcomes for disabled people and extending to other protected characteristics.
• Review current GDPR legislation working with AI owners (i.e., Big Tech companies) to enact a robust resolution to current data security concerns including to support disabled peoples’ continued access to assistive technologies.
• Awareness campaigns to support responsible use and highlight potential risks of AI including its use for advice and social support.

Skills, Accessibility and Digital Inclusion

• Tailor free-to-access digital skills and AI training courses (e.g., AI Skills Boost) targeting groups at risk of digital exclusion including people with learning disabilities, disabled older people, and economically disadvantaged/marginalised groups with delivery options for either online or supported in-person training sessions to maximise accessibility.

Access to Work and Employment Support

• Department for Work and Pensions to:
• Integrate AI and automation into current Access to Work processing through AI‑supported triage, case‑routing and case‑tracking including post-case evaluation to reduce delays and monitor value-for-money while retaining human-centred decision‑making.
• Design an assessor training package to improve consistency in assessment quality and information provided to disabled applicants of options available of assistive technologies and GDPR suitability.

2               Project outline

2.1               AI-EMPOWERED is funded by Edward Gostling Foundation (Charitable Registration No. 1068617).

2.2               The project has developed a rich evidence base collected in Q4 2025-Q1 2026 comprising a systematic literature review, survey of 1,029 employers, survey of over 1,300 disabled people, interviews with 65 disabled people, 47 employer representatives and over 40 stakeholders (academic/industry experts, non-government organisations, disabled people’s organisations, charities, government departments, trade unions, staff networks) and organisational case studies capturing lived experience of over 80 disabled people.

2.3               We adopt ‘disabled people’ in alignment with the UK disabled people’s movement and prevailing usage in UK policy, while also recognising administrative criteria for eligibility/receipt of statutory financial support. Our definition of disabled people includes physical or sensory impairment, chronic health concerns, learning difficulties, learning disabilities, long-term mental health conditions and neurodiversity, with this approach recognising frequency of co-occurrence of disability. Our method has supported data capture from all groups of disabled people within this definition.

3              Theme 2: AI adoption

3.1               Our evidence from a survey of 1,029 organisations conducted on our behalf by YouGov captures a broadly representative cross-section of the UK economy.

3.2               We find AI adoption remains in its infancy in many organisations and sectors: 29% of organisations reported not having adopted AI (Q4 2025), with adoption particularly low among micro (<10 employees) and small (10-49 employees) enterprises, and organisations in the private sector. We find 53% of micro and small organisations report not having adopted AI, compared with only 11% of large organisations (250+ employees).

3.3               Our data suggests a risk that micro and small enterprises could be left behind in AI-adoption. This suggests the need for Government intervention to develop a support package for businesses to provide minimum level AI awareness, leadership AI skills and support in sourcing secure AI solutions.

3.4               AI-integration within existing systems is, though, becoming increasingly common, reported by 37% of employers, as major technology companies and software/system providers have added AI functionality into existing platforms/systems, e.g., AI-integration via CoPilot into Microsoft 365.

3.5               Over 30% of organisations have put in place generative AI tools (e.g., Co-Pilot, Google Gemini, Chat GPT), for use by workers. Other forms of AI remain less common: AI-agents (22%), automated workflow management (19%), automated recruitment processes (9%) and physical task automation (8%).

3.6               AI use levels from our survey of disabled people in work indicate AI use is currently focused around: AI-integration within existing digital technologies such as CoPilot within Microsoft 365 (61% reported at least occasional use), content creation via generative AI (49%), AI-agents/assistants (43%), and AI-integrated assistive technologies (32%). Other uses are only reported by around 1-in-4 or fewer: task/workflow management, analytics, research and development, and physical task automation.[i]

4              Theme 4: Impacts on work and workers

4.1               Employment opportunities and experiences of work

4.2              Many workers will not experience direct engagement with AI while delivering a role: however, the worker experience may still be shaped by AI driven decisions/systems operating in the background, often without employees’ full awareness.

4.3               Understanding impacts for disabled people is important given the widening disability employment gap (29.7% Q2 2025 compared to 28.6% in Q2 2024) and increases in unemployment evidencing that disabled people are being more heavily impacted by the weakening UK labour market.[ii]

4.4               We heard from disabled people and stakeholders how disabled people could be at greater risk of job displacement owing to greater propensity to hold jobs more vulnerable to AI-mediated decisions that affect task automation, scheduling, performance evaluation and job security, in some cases without their knowledge and/or the protections afforded by formal disability recognition. However, the picture is complex with debates emphasising task-augmentation rather than whole-job replacement in many cases.

4.5               Benefits of AI adoption for disabled people currently look to outweigh these challenges and risks. Our employer survey indicates a net job creation benefit as increases in job opportunities were reported for entry level (14%), intermediate (12%) and senior (12%) roles, that are greater than reductions reported: entry level (10%), intermediate (9%) and senior (7%) roles. Organisations surveyed further reported that AI adoption has created greater employment opportunities for disabled people (14%).

4.6               Specific concerns were noted around risks of using AI systems in monitoring performance, including where disabled people have reasonable adjustments in place: ‘People are losing their jobs because HR technologies discriminate against disabled employees and don't take into account that the reason they didn't hit some target was because their manager said, as a reasonable adjustment, they didn't have to. But the process of determining whether you're performing will be measured in such a way by AI that you lose your job when you shouldn't.’ [Disabled people’s organisation representative]

4.7               Risks around quality of AI generated content were also highlighted. Almost 60% of disabled people surveyed reported that they do not trust content produced or tasks performed by AI. Our qualitative evidence emphasises risks including for people with learning disabilities and for disabled people with other protected characteristics including older adults and economically disadvantaged/marginalized groups.

4.8               AI-generated content is experimental resulting in risk of inaccuracies/hallucinations. While this is stated on AI user interfaces, content created may be trusted and accepted when containing significant errors. This could impact on performance in work contexts but also have wider implications in influencing people’s behaviours and leaving them vulnerable to misinformation.

4.9               A disabled worker in adult social care outlined in relation to vulnerable adults at the margins of the labour market: ‘I think there is some vulnerability in terms of whether you trust things that you're being told. I can see that the more vulnerable you are, the more problematic that might be.’

4.10           Wellbeing effects

4.11               Our employer survey suggests that, overall, use of AI may improve employee health and wellbeing (21% reported an improvement, compared to 4% a reduction). 

4.12           Our survey of disabled people highlights considerable uncertainty around effects of AI on the labour market and society with potential relevance to individual wellbeing as 64% noted concerns about AI taking over their jobs, while 55% held ethical concerns which reduced willingness to use AI.

4.13           AI can provide social support. Within our survey 37% of disabled people stated that they used AI for advice or social support. While disabled people we heard from were clear that AI is not a replacement/alternative for human interaction, some reported using it as a sounding board or log to manage their disability: ‘I use AI for advice with my disability and health condition, and I find that really invaluable… going into more detail than I would with anyone else, and thinking things through and identifying patterns in what would help to support me. I think that's a bit of a hidden function of AI that people are using it for.’ [Disabled employee]

4.14           Other disabled people reported using AI to ease anxiety/worries including communications sent by colleagues/managers where the disabled person found it difficult to interpret meaning or tone. This appears especially important for neurodiverse people but also those reporting mental health conditions, although we highlight the importance of clear signposting to professional support in AI systems given risks presented for mental health around AI dependency.             

4.15           Productivity and workplace inclusivity

4.16           Organisations in our employer survey reported significant productivity benefits from AI adoption: 40% reported increased productivity compared to only 2% a reduction. Employers recognised benefits for disabled people: 17% reported increased productivity compared to 3% a reduction (remainders report no change/don’t know)

4.17           Our interview and case study evidence provides understanding of benefits of AI and automation for disabled people, highlighting its potential in supporting enhanced performance and work inclusivity.

4.18           For individuals with physical impairments including dexterity or upper body mobility AI-integrated voice to text, and generative AI that produces draft content, was described as offering significant benefits in supporting work while reducing physical impacts of work.

4.19           Those with chronic conditions reported AI supporting more efficient routine and repetitive task completion including typing, and supporting them to take breaks and manage fatigue/pain including through tracking symptoms: ‘I use it to manage my health condition…I use it for dietary advice, I log situations that have happened and how I could [manage] it next time.’ [Disabled employee]

4.20           For neurodiverse people AI offers significant potential. AI supports more engaged communication with colleagues/clients/customers. AI is used to draft/refine written communications and interpret communications from others to support workplace relationships. AI tools support enhanced management of workflows and reduce task initiation difficulties through generation of initial ideas and structure in creating documents/outputs.

4.21           For the visually impaired AI-integration into screen reading tools has enhanced accuracy while emerging AI-integrated wearable technologies can provide information to the wearer on supporting human interaction and independence (e.g., commuting/travel).

4.22           For D/deaf and hard of hearing people AI-integration into auto-transcription improves quality and more reliable communications, meeting summaries and documents, while AI British Sign Language tools are being developed.

4.23           Advancements in voice-to-text and text-to-voice also support AI-mediated interaction between groups of disabled people, for example, visual and hearing impaired. This should support employers in feeling more confident in employing diverse groups of disabled people.

4.24           Evidencing the whole life potential of AI to support disabled people in(to) work, we heard from people with learning disabilities how using AI-integrated tools that help them manage their time and provide guidance (e.g. travel), management of finances, and other aspects of life, can support them to engage with paid work. 

4.25              Recruitment and HR processes

4.26              AI-integrated recruitment systems that screen, shortlist and even conduct interviews offer significant potential for efficiency gains in hiring processes.

4.27              Our evidence shows risks for disabled people that may perpetuate or exacerbate existing societal biases through automated HR systems that screen, shortlist and perform hiring processes.  In our survey of disabled people 63% held concerns that use of AI in recruitment/hiring processes discriminates against them.

4.28              A disabled people’s organisation representative offered an example: ‘So you graduate from [prestigious University for the Deaf], so you've got two things that happen here. One is that the machine learning has taught this AI tool that deaf people don't get jobs. Second, you're excluded because it has not been taught that [University] is credible. So, you're discarded because you're deaf and you're at a prestigious college for the deaf.’

4.29              Bias and potential discrimination in recruitment is a concern highlighted, but one that is difficult to accurately measure. The lack of transparency in AI systems in how they make decisions in screening, shortlisting and interviewing leaves a high degree of uncertainty about potential impact on opportunities for disabled people, requiring controlled trials to accurately measure impacts.

5                      Theme 5: Skills, education and transitions

5.1               Many widely used AI tools including generative AI (CoPilot, Chat GPT, Google Gemini) have low entry skill requirements for basic use. AI-integration in existing systems and tools has similarly been enacted to support low skill requirements.

5.2               AI was noted as having a great levelling potential, offering ways to personalise experiences to individual needs: ‘You can set it up so if you want it to speak to you in a certain way and style, you can.’ [Disabled employee]

5.3               Almost all the disabled people we interviewed outlined how they had learned at least basic AI skills through direct use of systems/tools.

5.4               Of the disabled people we surveyed 30 percent reported having received formal AI training through their employer or other sources. Disabled people interviewed reported seeking training through free-to-access social media platforms, e.g., YouTube, where workplace training was not offered.

5.5               AI was reported as offering benefits through filling existing employee skills gaps and as a training resource: ‘When I transitioned [to new role], there was a little bit of a skills gap and AI just helped make that gap close… being able to use it as a training tool.’ [Disabled employee]

5.6               Some disabled people reported accessibility difficulties as systems did not respond in intended ways or require specific approaches to generate intended results: ‘We've done a bit about prompt engineering in our Department, there's different methods and weird like bias and stuff in the way that you should tell it to structure a result.’ [Disabled employee]

5.7               Gaps in understanding and difficulties in effective use in a limited number of cases had resulted in disabled people discontinuing use.

5.8               AI-integration into assistive technologies in some cases supported enhanced functionality but in limited cases creates challenges in use or results in specific assistive technologies being withdrawn by employers on grounds of AI breaching GDPR (see Theme 6).

5.9               Disconnects were reported between employers and Government support provided by Access to Work with AI-integrated assistive technologies recommended that did not meet GDPR requirements.

5.10           Disconnects also resulted in training in assistive technologies provided through Access to Work not being provided and the assistive technology therefore not being used.

5.11           Wider digital skills gaps remain a significant barrier to employment for people with learning disabilities, older disabled people, and economically disadvantaged/marginalized groups.

5.12           Although outside of the direct scope of AI-EMPOWERED, embedding AI literacy in the education system is essential.

5.13           Our evidence highlights a need to avoid potential widening of digital skills gaps that could perpetuate negative employment outcomes for disabled people, requiring tailored basic post-education digital skills and AI training to provide basic literacy in AI including generative AI and AI-integrated digital tools/systems with delivery options for either online or supported in-person training sessions to maximise accessibility.

6                      Theme 6: Government Strategy, Regulation & Rights

6.1               In addition, to evidence already provided in response to other themes, we highlight here implications of AI-integration for GDPR and availability of assistive technologies for disabled people, and the centrality of disabled people’s voice.

6.2               AI-integration into assistive technologies has in some cases resulted in systems/tools used prior to AI-integration breaching GDPR rules. Assistive technologies have in turn been withdrawn from approved lists in organisations, leaving gaps in support for disabled people as organisations and schemes such as Access to Work struggle to keep up with the pace of change.

6.3               ‘There was no prior warning, I just lost access completely overnight and then was told we won't renew them because of GDPR … I ended up having to be signed off work with stress and anxiety because I'd built a workflow and a system that made my life so much easier and to have that withdrawn completely overnight, threw everything into chaos for me and I didn't know how I was going to cope and manage.’ [Disabled employee]

6.4               A visually impaired interviewee told us that they were concerned about risks of AI solutions becoming a default, using the example of a valued and necessary support worker potentially being replaced: ‘The DWP will go, hold on, we’re giving these blind people all this funding… well that blind person, she doesn’t need a support worker because she can just have an AI note taker’.

6.5               The importance of giving a voice to disabled people around the future of employment support and labour market and technology change was emphasised across our evidence: ‘I don't think personally that they have the right people thinking of ideas for disabled people. It needs to be, ‘actually, I live like this, this is beneficial to people’. And I don't think that is something that is fully realised. I think there is sort of an endemic issue that there isn't a realisation of how [disabled] people actually function.’ [Disabled employee]

6