How to build accessible AI: Insights from industry leaders
Robin Christopherson | 15 Sep 2026Last week AbilityNet had the privilege to co-host an AI and accessibility roundtable with the team at Skyscanner, bringing together accessibility leaders from a wide range of household-name organisations across sectors including financial services, retail, media, technology and the public sector. The discussion was held under Chatham House rules, which meant people spoke openly about both the opportunities they are pursuing and the concerns that are keeping them awake at night.
As someone who spends much of my time working with organisations on digital accessibility, and as a blind person who uses AI every day, I came away with two strong impressions.
The first is that AI is now moving from experimen
tation into operational reality. The second is that accessibility has become one of the most important tests of whether organisations will adopt AI successfully and responsibly.
What made this session so valuable was its honesty. This wasn’t a room full of people claiming to have solved the challenges. It was a room full of experienced practitioners comparing notes, sharing lessons and trying to work out what good looks like in a landscape that is changing week by week.
A useful reminder: nobody has all the answers
One of the most reassuring aspects of the discussion was hearing just how many organisations are still working through similar questions.
There is a tendency to assume that larger organisations have already figured everything out. The reality is far more nuanced. Teams are experimenting with AI in different ways, governance frameworks are still evolving, and accessibility specialists are often trying to support an unprecedented rate of change.
Several participants described a situation where AI adoption is advancing faster than established processes can comfortably accommodate. New tools are appearing, developers are using AI-assisted workflows to create code at scale, product teams are exploring automation opportunities, and accessibility professionals are being asked to provide guidance in an environment that shifts almost daily.
That may sound challenging, but there was an encouraging undertone throughout the conversation. Organisations are sharing ideas, learning from one another and increasingly recognising that accessibility needs to be part of the AI conversation from the beginning.
The biggest risk is not the one many people expect
When people talk about AI risk, there is often a focus on dramatic future scenarios. The discussion in the room, however, repeatedly returned to a much more immediate concern; that the biggest risk today may be false confidence.
AI can generate accessibility advice, code, audit findings, remediation recommendations and policy documents that sound entirely plausible. Sometimes they are excellent. Sometimes they are subtly wrong. Often they neglect to mention those areas that were beyond their agents’ abilities to address. The challenge is that the output often carries an air of authority regardless of its accuracy.
Several participants highlighted examples where AI-generated accessibility guidance appeared convincing enough that non-specialists would be unlikely to question it. In some cases, even experienced practitioners would need to take a closer look.
That observation resonated strongly with my own experience. AI can be an extraordinarily capable assistant. It can help me research, draft, analyse and explore ideas at a speed that would have seemed remarkable only a few years ago. At the same time, I never assume it is right simply because it sounds confident.
Where the output is critical, human judgement remains essential.
Accessibility teams are already stretched
Another recurring theme was capacity. Many accessibility teams remain remarkably small compared with the scale of the organisations they support. Participants shared examples of relatively small specialist teams providing guidance across hundreds of services, products or customer journeys.
One team shared how they are now seeing around 7,000 submissions of new code every week. This new AI-driven level of development is completely unprecedented, and impossible to monitor in its entirety by what still remains a largely human-driven accessibility assurance process. Surely automation is called for.
This helps explain why AI has generated such interest within the accessibility profession. The attraction is not simply efficiency for its own sake. It is the possibility of extending expertise further, helping teams identify issues earlier, surfacing potential problems at scale and making existing specialists more effective.
Accessibility professionals have always had to think carefully about where they invest their time. AI has the potential to help with that challenge, particularly when it can automate repetitive tasks and allow experts to focus on the problems that require deeper analysis and judgement.
Scaling good practice through AI
One insight appeared repeatedly throughout the discussion; the organisations seeing some of the most meaningful results are not necessarily using AI to create entirely new approaches. Instead, they are using it to help people adopt existing good practice more consistently.
Design systems featured prominently in the conversation. When accessible components are available, trusted and widely used, they allow good accessibility practice to scale across large organisations. AI is proving particularly useful in helping teams migrate older products, increase consistency and encourage adoption of those established patterns.
The long-term vision discussed by some participants was even more ambitious: accessibility becoming embedded within the generation process itself, with tools producing accessible outcomes by default rather than relying on defects being identified and fixed later.
Whether that future arrives quickly remains to be seen, but it is an exciting prospect.
Better AI starts with better foundations
A particularly practical takeaway concerned how organisations are training and configuring AI systems.
Generic models tend to produce generic answers. The most effective approaches discussed at the roundtable involved grounding AI in trusted internal resources such as design systems, accessibility guidance, coding standards and organisational knowledge. 
In other words, organisations are discovering that AI becomes significantly more useful when it learns from the standards and expertise they already trust.
That principle applies far beyond accessibility. It is equally relevant in sectors such as financial services, where organisations need AI outputs that align with regulatory requirements, customer vulnerability strategies and established internal policies. The lesson is straightforward: if you want trustworthy outcomes, start with trustworthy inputs.
Accessibility at scale cuts both ways
Throughout the day there was a shared sense of optimism, but it was a careful optimism.
Given carefully applied constraints and training, AI can help organisations apply good accessibility practices more widely and more consistently than ever before. It can accelerate testing, support developers, improve workflows and enable forms of personalisation that were previously difficult to imagine.
As a blind person, I find some of these possibilities particularly exciting. The idea of interfaces that adapt more intelligently to individual needs has real potential. Rather than expecting people to conform to a single experience, technology could become much better at accommodating diverse requirements from the outset.
Yet the same technologies that can spread good practice rapidly can also spread poor practice rapidly. An inaccessible pattern recklessly replicated thousands of times remains inaccessible. The scale works in both directions.
That is why accessibility needs to remain closely connected to disabled people's lived experiences. Several participants discussed the use of ‘synthetic personas’; AI-generated simulations of diverse users. These tools may have genuine value, especially in helping teams explore scenarios earlier in the design process, but they cannot replace engagement with real, live, disabled users.
There is still no substitute for understanding how people actually experience technology. Just as with a comprehensive audit of your code, for the time being at least, humans remain at the heart of the process.
What I took away
Perhaps the strongest impression from the event was the willingness of participants to learn from one another. The discussion was practical, candid and refreshingly grounded. People shared successes, highlighted mistakes, questioned assumptions and explored ideas that are still taking shape.
Nobody suggested that AI would remove the need for accessibility expertise. If anything, the opposite appears to be true. As AI becomes embedded across organisations, the importance of accessibility knowledge, governance, testing and lived experience is only increasing.
AI is already changing how accessibility work is done, but it has yet to change the fundamental reason why that work matters.
How AbilityNet can help
Many organisations are now asking similar questions. How should accessibility teams engage with AI initiatives? How can AI be used responsibly within accessibility programmes? What governance, testing and assurance approaches are most effective? And how do you ensure disabled customers and colleagues are not left behind as adoption accelerates?
At AbilityNet, we help organisations answer those questions through accessibility strategy, audits, user testing with disabled people, training, consultancy and AI-focused accessibility guidance.
Explore AbilityNet's accessible AI integration and consultancy services or speak to our experts about an AI feature, customer journey or deployment that you want to assess.
Talk to us about accessible AI integrations
More from Robin Christopherson on AI and accessibility
Five questions to ask about every AI deployment
- Published 8 Sep 2026. AI could generate inaccessible websites, apps and content at such scale that organisations end up with an even bigger accessibility backlog than before.
Accessible AI: Who is responsible when AI acts on our behalf?
- Published 2 Sep 2026. As AI takes on more tasks, accessibility becomes more important than ever.
Why accessibility is one of the most important AI procurement decisions you can make
- Published 24 Aug 2026. Inaccessible AI can create costly remediation, governance risks and poor customer experiences.
The hidden cost of AI bias: Why disability may be the ultimate test of trustworthy AI
- Published 17 Aug 2026. As AI adoption grows, disability is emerging as a critical test of trustworthy AI.
The EU AI Act has a message for accessibility teams: get involved earlier
- Published 10 Aug 2026. The EU AI Act's requirement for human oversight highlights the need to involve accessibility experts and disabled users much earlier in AI development.
Using AI tools to drive an accessible dev process
- Published 05 Aug 2026. How do we move beyond AI-generated code and start using AI to reduce accessibility debt before it reaches production?
AI could make financial services more inclusive. But only if accessibility is part of the infrastructure
- Published 27 July 2026. Why accessibility must be treated as core infrastructure if AI is to improve financial inclusion.
The trust gap: why AI's edge cases matter to customer loyalty
- Published 30 June 2026. Why accessibility and edge cases are critical to building trust in AI-powered services.
The delicious irony at the heart of trusting AI
- Published 17 June 2026. A personal reflection on trust, accessibility and the unexpected ways AI can both help and hinder.
Executive briefings for senior leaders
Robin offers invite-only executive briefings exploring AI, accessibility, governance, customer trust, regulation and digital inclusion.
