Episode 97: Building a More Inclusive Future: How AI is Transforming Accessibility

Ioana Tanase02/20/2025

The future of accessible AI isn’t just about creating new technologies — it’s about ensuring technologies serve everyone regardless of ability. One particularly promising application of AI in accessibility is its potential to support healthcare services. As the global demand for healthcare systems continues to outpace the availability of practitioners, innovative AI solutions are emerging to help bridge this critical gap.

In this episode of the #shifthappens podcast, Ioana Tanase, Microsoft’s Inclusive AI and Accessibility Program Manager, shares insights into how AI can transform accessibility and create more inclusive digital experiences. Her perspective offers a unique glimpse into the intersection of AI development and accessibility, revealing the unprecedented opportunities and critical responsibilities that come with this technological innovation.

The Current State of Digital Accessibility

Ioana presents a stark reality check: “97% of the world’s websites are not accessible.” This reality affects millions of people worldwide, making routine tasks like booking travel arrangements unnecessarily complicated for individuals with disabilities.

However, through her work, she has witnessed firsthand how AI can serve as a powerful tool for accessibility. Microsoft’s AI for Accessibility program provides grants to organizations developing innovative AI solutions for people with disabilities. These initiatives range from assistive technologies to AI-powered accommodations that make digital experiences more accessible.

Ioana, who is dyslexic herself, shares personal examples of how AI has transformed her work experience: “I do need to read a lot of research papers... and being able just to reason on top of a document and ask questions is massive.” She envisions AI agents playing a crucial role in creating more equitable digital experiences.

The Role of Meta-Prompting and Agents in Accessibility

Looking toward the future, Ioana describes a scenario where an AI agent could assist a blind user with travel arrangements: “How amazing would it be for my colleague just to be able to pull in an agent that has his information, knows his preferred traveling paths, knows the preferred company that he does the taxi transfer with, knows what his loyalties with this hotel chain or the other, [and the AI agent] does all of that automatically.”

Nevertheless, she emphasizes a vital caveat: “I don’t think AI should be an excuse for people not to make their technology or website accessible, but it definitely can help with those gaps and ensure that the experience is equitable as opposed to right now — which it just isn’t.”

As AI becomes more integrated into accessibility solutions, privacy and safety considerations become paramount. Ioana’s work on responsible AI mitigation, including meta-prompting (also known as safety system messaging), helps address potential copyright, privacy, and content safety-related harms.

Meta-prompting serves as a safety mechanism, providing clear instructions to AI models about appropriate behavior and limitations. As Ioana describes: “All of these models have instructions of what to do and what not to do. So basically, you’re telling the model, this is what you are. This is how you should engage with a user, and this is what you absolutely should not do.”

She explains the multi-layered approach to AI safety: “With any AI system, there are multiple levels of mitigation. It starts with the model itself — the model can come ingrained with certain safety limits, like copyright or the data that is being trained on, and so forth.”

Use CaseScenario: Healthcare and AI

A promising application of AI in accessibility is its potential to support healthcare services. Ioana highlights several areas where AI is making an impact:

• Responsible representation of mental health conditions in AI systems
• Design considerations to promote healthy digital habits
• Innovative solutions to address the global shortage of healthcare practitioners

She shares an example of successful implementation with one of the organizations Microsoft works with through the AI for Accessibility program. IWill, based in India, created a cognitive behavioral therapy (CBT) model, where if you have low to moderate forms of anxiety or stress, you can get real-time CBT treatment through a particular chatbot. Ioana explains, “They took the Azure OpenAI model but grounded it into actual data. The development was guided by practicing psychologists, and the model incorporates actual patient treatment data and outcomes.”

This targeted approach helps address a critical challenge in mental healthcare delivery. As Ioana notes, “It makes it easier for people to get treatment, and it also puts less pressure on practicing physicians to be able to work on high cases of anxiety or stress as opposed to the low to moderate ones.”

However, Ioana stresses the importance of maintaining appropriate boundaries and human oversight in healthcare applications. “Your user should be aware of the type of system they're interacting with, ensuring that the human is still the one that can control the experience. And decisions are not being made automatically without that human in the loop,” she adds.

When developing AI systems, privacy and safety considerations become even more critical. Microsoft's approach includes:

• Strict data privacy protocols to protect sensitive information
• Clear boundaries on what AI systems can and cannot do
• Regular testing and validation of AI responses
• Continuous monitoring for potential biases or harmful outputs

Call to Action for Technology Stakeholders

Transparency and human agency are key elements of responsible AI development. Ioana concludes with specific recommendations for different stakeholders in the technology industry:

• For Developers: Closely review your products and ensure you have implemented responsible technology development principles. “We absolutely need to ensure that everything we’re producing is responsible. You can add whatever flavor you want, but it shouldn’t put people at risk.”
• For Business Leaders: Think of ways to support your workforce’s productivity with AI tools and ensure you get the best out of your data. “Consider the AI opportunity because, honestly, having used this technology for so long, I know my productivity has changed.”

While the current state of digital accessibility remains challenging, AI presents unique opportunities to bridge existing gaps and create more equitable solutions. This requires an ongoing commitment from developers, business leaders, and technology companies to prioritize accessibility and inclusion in AI development.

Looking Ahead

As organizations continue to develop and deploy AI solutions, it will be crucial to keep accessibility at the forefront of design and implementation to build a truly inclusive digital world.

Successful integration of AI into accessibility solutions depends on careful consideration of privacy, safety, and ethical implications. As Ioana’s insights reveal, the key lies in developing AI systems that enhance human capabilities while maintaining human agency and control.