Can AI open doors that technology has left closed?
Imagine a person who has excellent ideas but finds written forms difficult to complete. Another person may communicate more easily with visual prompts than spoken instructions. Someone else might want to enter employment but struggle with an application process that does not accommodate their access needs. A disabled artist might have a strong creative vision but need support with editing, transcription, organisation or presenting their work.
Artificial intelligence could help with some of these barriers. It can convert speech to text, summarise information, support translation, generate captions, organise tasks and help people experiment with creative tools. Yet none of these features automatically makes a system inclusive. Speech recognition can misunderstand someone’s voice. A summary can remove important details. A chatbot can provide incorrect advice. An automated recruitment system can reproduce unfair assumptions about who makes a good employee.
The important question is not whether AI can do impressive things. It is whether it helps people achieve outcomes they value, with control over their choices, privacy and dignity.
What does inclusive innovation actually mean?
Inclusive innovation starts by involving the people who will use a product or service in deciding what problem needs solving. It does not treat disabled people as a single group with identical needs. Disability, neurodivergence, communication preferences, sensory access, mobility, digital confidence and personal circumstances vary widely.
A tool that works well for one person may be unhelpful to another. A person may prefer plain text, audio, symbols, video, a human conversation or a combination of formats. Some may welcome automation; others may want technology to remain in the background while a trusted person supports them. The choice should not be made on their behalf without consultation.
Inclusive design also asks whether the problem is technological in the first place. Sometimes the best solution is a clearer form, a quieter room, a flexible appointment, a better staff procedure or more time. AI should not be added merely because funding or publicity favours new technology.
Where could AI make a practical difference?
Communication and understanding: AI-assisted transcription, translation and plain-language summaries may help people access information. Outputs must be checked, particularly when they concern safeguarding, legal rights, benefits, health or important decisions.
Learning and employment: AI can help people practise interviews, organise tasks, draft applications and explore learning materials. It should support the individual rather than automatically judge their suitability for work or training.
Creative participation: Speech-to-text, captioning, sound tools, image description and accessible production workflows may help more people take part in music, film, storytelling and design. The participant should remain the author of their ideas and retain meaningful choices over the final work.
Everyday organisation: Optional reminders, step-by-step instructions and personalised planning tools may reduce the effort involved in some tasks. They should be adjustable and should not create a new system that people are expected to obey without question.
Access to services: AI may help organisations make information easier to find or offer it in different formats. It should not become a barrier between a person and a human decision-maker, nor should a person be forced to disclose unnecessary personal information simply to obtain support.
What could go wrong?
One risk is inaccurate assistance. A system may confidently produce a wrong answer, misunderstand a person’s communication or omit a crucial qualification. If someone relies on it to understand an appointment, a support plan or an employment opportunity, the consequences may be significant.
Another risk is the collection of sensitive information. Disability-related information can be highly personal. Conversations, voice recordings, support notes, behaviour observations and details about someone’s needs should not be uploaded to public AI tools without carefully assessing the purpose, lawful basis, privacy terms and risks. A project should collect the minimum information necessary and explain clearly who can access it and how long it will be retained.
Bias is also important. If an AI system has been developed or tested using data that does not adequately represent disabled people, it may work less reliably for them. A recruitment tool, for example, could penalise an applicant whose communication style differs from the patterns it has learned to expect. A system that assumes one way of speaking, learning or behaving is the norm can turn an existing barrier into an automated one.
The Information Commissioner’s Office explains that AI systems must be considered in relation to data protection fairness and discrimination risks. It also emphasises that compliance with data protection law does not, by itself, guarantee compliance with equality law. Organisations should assess the real effects of a system and consider whether AI is an appropriate solution at all. Read the ICO’s AI guidance.
Three questions that should shape every project
- Was the solution designed with the people who will use it? Disabled and neurodivergent participants should help define the problem, test the product and decide what counts as success. Participation should be accessible and appropriately compensated where possible.
- Can the person say no? Users should understand what the tool does, what information it processes, its limitations and how to request human help. They should not be pressured into using it or punished for preferring another method.
- Does it improve an outcome that matters? Measure whether people can understand information more easily, complete a chosen task, participate more independently or access opportunities that were previously difficult. Do not count downloads or log-ins as proof of inclusion.
From idea to service: three possible solutions
The first option is an inclusive AI access service that helps organisations make their information more usable. Rather than developing a new AI model, it could combine plain-language editing, accessible document design, optional AI-assisted summaries, captions and human review. The service would work with users to identify barriers and check whether revised materials genuinely help.
The second option is a supported employment and skills programme. Participants could use AI to practise interviews, explore job roles, draft applications or plan tasks, with a trained human adviser available. The programme would measure whether people gain confidence, access interviews, obtain suitable work or remain in employment. It would not use AI to decide who deserves support.
The third option is a creative technology lab for disabled and neurodivergent artists. Participants could experiment with music production, filmmaking, storytelling, visual design and accessible editing tools. They would choose what they want to create, retain agreed rights to their work and have a say in how any output is shared. The programme could also produce a practical guide for arts organisations wanting to improve accessibility.
What is the business opportunity?
There is a potential market for organisations that need accessible content, inclusive training and practical support to introduce technology responsibly. Potential customers include charities, employers, education providers, arts organisations, local services and training organisations. They may need help selecting tools, adapting workflows, training staff and evaluating outcomes.
A small provider does not need to compete with major technology companies by creating another general-purpose AI assistant. It could instead offer a specialised service: understanding the user’s needs, choosing suitable existing tools, configuring accessible workflows, training people and evaluating whether the solution works.
The differentiator would be trust and delivery. The service should be clear about what technology can and cannot do, involve people with lived experience in meaningful roles, protect personal information and make human support available when it is needed.
For AM-MediaSpace, a possible starting point is a creative technology lab where disabled and neurodivergent participants explore music, film and storytelling. This would connect inclusion with creative work without assuming that every participant wants the same activity or that disability should define their identity. The model could later be adapted for other organisations, provided the evidence supports doing so.
What funding could support the work?
Funding should follow the project’s purpose rather than the other way round. A project designed to improve disabled people’s access to employment may fit different funding criteria from a creative arts programme or a commercial accessibility product. A fund that supports employment outcomes may not pay for a general music workshop, and a creative grant may not cover software development or ongoing business costs.
One relevant UK programme is the Department for Work and Pensions’ Pathways to Work Innovation Fund. Its published guidance describes funding to test new approaches that help disabled people and people with long-term health conditions move towards, enter, stay in or progress in work, while building evidence about what works. The first round offered up to £10 million, but its application deadline was 26 October 2026, which has now passed. The guidance says a further round of up to £50 million is expected to open in early 2027. This is a possible future route for a genuinely employment-focused innovation, not an available grant for every inclusion project. Read the official guidance.
Innovate UK also lists BridgeAI Deployment Catalyst competitions for UK-registered small and medium-sized businesses deploying AI. The published opportunities include separate themes for decision-critical systems, operational systems and assurance/compliance, each with up to £5 million available across the competition. The listed closing date is 8 December 2026. A disability-related idea would need to match the exact competition brief, meet its eligibility and deployment requirements, and demonstrate a credible technical and commercial case. The overall fund amount is not the amount any one applicant will receive. Check current UKRI AI opportunities.
Other routes may include charitable trusts, local authority or community partnerships, arts and creative funding, research collaborations and employer-funded pilots. These should be checked individually for eligibility, match funding, eligible costs, safeguarding requirements, accessibility expectations and whether a commercial organisation can apply directly or needs a suitable lead partner.
How should a project prove that it works?
A pilot should begin with a clear baseline. If the goal is to help people understand information, record how accessible the original information is and ask participants what they find difficult. If the goal is employment, agree measures such as completed applications, interview opportunities, participant-defined goals and job retention. If the goal is creative participation, measure whether people can access the tools, make their chosen work, influence decisions and decide how the work is shared.
Evaluation must not reduce people to a score. Feedback should be available in accessible formats, and participants should be able to decline questions. Where personal or sensitive data is not needed, do not collect it. If a project uses AI to process personal data, assess the data protection implications before the pilot and involve a qualified adviser when necessary.
It is equally important to record negative results. Perhaps a speech tool works poorly for some participants. Perhaps a simpler non-AI interface performs better. Perhaps people value human support more than automated prompts. Those findings are useful. A responsible project is allowed to conclude that a tool should not be used.
What if the project fails?
Set stop conditions in advance. Pause the pilot if participants cannot give informed consent, personal data is exposed, the tool repeatedly produces harmful or misleading results, or people feel pressured to use it. Review complaints with people who have appropriate authority and experience, and do not rely on the AI system to investigate its own failures.
Financially, avoid building a custom platform before testing the need. Start with a limited pilot using established tools and accessible alternatives. Keep the project’s software, data and processes as portable as possible. If the service cannot demonstrate measurable value, revise it or stop rather than continuing simply because money has already been spent.
Conclusion: inclusion is measured by the doors people can actually open
AI can be useful when it gives people more ways to communicate, create, learn, work or participate. But inclusion is not achieved by adding a chatbot to an inaccessible service or describing a product as innovative. It depends on whether people can use it, whether it respects their choices and whether it improves something that matters to them.
The best projects begin with listening. Disabled and neurodivergent people should help set the priorities, test the technology and judge the outcomes. Human support must remain available, privacy must be protected and no one should be treated as less capable because an algorithm misunderstands them.
The commercial opportunity is to help organisations turn inclusive intentions into workable services. The public benefit is to make participation and opportunity more accessible. Those aims can reinforce each other, but only if the project measures results honestly and remains willing to change direction.
The first step is simple: choose one barrier, speak to the people experiencing it, test whether AI is genuinely useful, and measure the result before expanding. Technology may help open a door, but people must retain the power to decide where they want to go.
Comments