Agentic Artificial Intelligence and the Future of Digital Accessibility

Agentic Artificial Intelligence and the Future of Digital Accessibility

Digital accessibility has traditionally focused on making websites, documents, and applications usable for people with disabilities through standards such as the Web Content Accessibility Guidelines (WCAG). A major technological shift is now underway. Artificial intelligence systems are evolving from passive tools that respond to commands into agentic systems capable of making decisions, completing multi-step tasks, and acting on behalf of users. Agentic AI may become one of the most significant developments in accessibility since the rise of the internet itself.

This chapter introduces agentic artificial intelligence and examines how autonomous AI systems may change digital accessibility. It explores how AI may expand access, personalize interactions, and automate accessibility supports, while also creating new risks related to bias, privacy, accountability, and user control. The chapter encourages readers to think critically about the future of accessibility in AI-driven digital environments.

Short Video: AI and Accessibility

The following short video explains how artificial intelligence can support accessibility and highlights the importance of responsible design.

Introduction: Accessibility at an AI Turning Point

Digital accessibility has often focused on ensuring that users can access information and complete tasks within systems that may not have been designed with disability access in mind. Agentic AI changes this model. Instead of only providing tools that help users interpret interfaces, AI systems may increasingly act as intermediaries that help users navigate digital environments, translate information, and complete complex tasks.

This shift suggests a new accessibility model in which technology adapts to the user more actively. At its best, agentic AI could support greater independence, personalization, and flexibility. At its worst, it could automate exclusion, reduce user control, and create new barriers that are harder to detect.

What Is Agentic AI?

Agentic AI refers to artificial intelligence systems that can perform goal-directed actions with a degree of autonomy. Unlike traditional software that waits for direct input at each step, agentic systems may be able to:

• Understand goals
• Plan actions
• Execute tasks autonomously
• Learn from outcomes
• Interact across multiple digital environments

Examples may include AI assistants that complete online forms, navigate websites, schedule appointments, gather information, or communicate with other systems on behalf of a user.

From Assistive Technology to Autonomous Assistance

Historically, accessibility has depended on specialized assistive technologies that help users work within existing digital systems.

Traditional Accessibility Model

• Screen readers
• Speech recognition software
• Alternative input devices
• Captioning systems

Agentic Accessibility Model

Agentic AI introduces a new paradigm in which technology becomes a more active accessibility partner. Instead of only helping users interpret content, AI agents may:

• Interpret inaccessible interfaces
• Convert content formats automatically
• Work around digital barriers in real time
• Personalize interaction methods

For example, an AI agent might navigate a poorly designed website, extract the needed information, and present it in a simpler and more accessible format.

Opportunities for Digital Accessibility

Personalized Accessibility

AI systems may adapt to individual needs and preferences rather than offering the same accessibility support to every user. Personalization may include:

• Reading complexity levels
• Preferred interaction methods
• Cognitive load tolerance
• Sensory sensitivities

This could shift accessibility from standardized compliance toward more adaptive support.

Real-Time Accessibility Remediation

Agentic systems may automatically improve inaccessible content by:

• Generating image descriptions
• Repairing inaccessible PDFs
• Creating captions and transcripts
• Simplifying complex language
• Reformatting layouts for readability

These capabilities may reduce the need for manual retrofitting, although they should not replace accessible design from the start.

Expanded Independence

AI agents may help users independently manage tasks that are often difficult in inaccessible digital environments, such as:

• Completing bureaucratic processes
• Managing educational platforms
• Using employment systems
• Navigating healthcare portals

This represents a shift from accommodation toward greater digital autonomy.

Accessibility in Education

In higher education, agentic AI may support:

• Adaptive learning environments
• Automatically accessible course materials
• Intelligent tutoring systems
• Real-time lecture accessibility supports

Students may increasingly interact with course systems through personalized AI intermediaries that help them access content and complete academic tasks.

Risks and Accessibility Challenges

Despite its promise, agentic AI introduces new risks that educators, institutions, and designers must consider carefully.

Algorithmic Bias

AI systems trained on incomplete or biased datasets may:

• Misinterpret disability-related behaviors
• Produce inaccurate captions or descriptions
• Exclude marginalized users

Accessibility failures may become automated and scaled across systems.

Loss of User Control

Autonomous systems raise important questions about control and accountability. For example:

• Who controls accessibility preferences?
• Can users override AI decisions?
• What happens when an AI system acts incorrectly?

Accessibility should empower users, not replace their agency.

Over-Reliance on AI Fixes

Organizations may be tempted to rely on AI overlays or automated remediation instead of building accessible systems from the start. This creates a risk that accessibility becomes reactive rather than foundational. Accessible design must remain a core responsibility.

Privacy and Surveillance Concerns

Agentic accessibility systems may require sensitive personal information, including:

• Disability-related information
• Behavioral patterns
• Communication styles
• Personal preferences

Ethical implementation requires clear consent practices, strong privacy protections, and transparent data governance.

Rethinking Accessibility Frameworks

Current standards such as WCAG are primarily designed to evaluate interfaces, content, and interactions that can be inspected in relatively stable ways. Agentic environments introduce new questions that are harder to measure with a static checklist.

Future accessibility frameworks may need to consider:

• Decision transparency
• User override capability
• Fairness in personalization
• Continuous accessibility performance

In AI-driven environments, accessibility may become an ongoing process of monitoring, testing, and revision rather than a one-time compliance task.

Human-AI Collaboration and Inclusive Design

The future of accessibility is unlikely to be fully automated. Strong accessibility practice will still require:

• Human-centered design
• Input from disability communities
• Ethical AI governance
• Continuous evaluation

People with disabilities should remain active participants in shaping agentic systems, not simply end users who experience the consequences of design decisions made by others.

The Role of Higher Education

Colleges and universities play an important role in shaping accessible AI futures. Institutions can:

• Teach responsible AI development
• Model accessible AI adoption
• Evaluate emerging technologies critically
• Prepare students for inclusive digital workplaces

Digital accessibility education now increasingly intersects with AI literacy, ethics, and public responsibility.