Publication Accessibility

A11yJam Roundtable Report: AI Tools for Accessibility

How can governments use artificial intelligence (AI) to advance digital accessibility? During this roundtable, Andrew Merluzzi, AI Innovation + Incubation Fellow at the Beeck Center for Social Impact + Innovation, shared some applications of AI in digital accessibility. Attendees discussed which applications are most promising, which seem too difficult or risky, and how their agencies could start experimenting with AI.

Author: Kirsten Wyatt
Published Date: Aug 5, 2026
Last Updated: Aug 5, 2026

AI Tools for Accessibility

On April 15, 2026, the Beeck Center hosted Accessibility Jam, which featured four interactive roundtable discussions. This is an anonymized report out of the discussion, showcasing key insights, questions, and topics. Roundtable sessions were only open to government employees, with the goal of fostering open and honest discussions about the roundtable topics.

Show & Tell Demo Video

Key Roundtable Insights

  1. Capacity constraints, not awareness, are the primary barrier

Accessibility is widely understood as important, but progress is constrained by limited staff, competing priorities, and insufficient time to implement at scale.

  1. The “PDF problem” remains the single largest operational burden

High-volume, legacy PDF remediation continues to dominate accessibility work, crowding out more sustainable approaches like structured, web-first content.

  1. Accessibility is fragmented across tools, teams, and standards

Inconsistent understanding and implementation across platforms (web, dashboards, SharePoint, reporting tools) create gaps, duplication, and uneven user experiences.

  1. AI interest is high, but evidence and clarity are low

There is strong curiosity and early experimentation with AI for accessibility, but limited shared evidence on what works, what fails, and where to invest next.

  1. Human oversight remains essential to trustworthy accessibility

AI can accelerate repetitive tasks, but risks of inaccuracy (e.g., captions, alt text, plain language) reinforce the need for human review and judgment in final outputs.

  1. The field is navigating uncertainty around policy, practice, and the future of AI

Open questions ranging from regulatory scope (e.g., email) to agentic design systems highlight a broader need for guidance, shared standards, and practical playbooks.

Roundtable Questions & Summary Answers

What are your AI and accessibility challenges?

  • Competing priorities limit capacity to implement accessibility work
  • Vendor compliance and accountability remain inconsistent
  • Lack of a shared understanding of accessibility across teams
  • Limited time and resources constrain progress
  • Large-scale PDF remediation efforts, sometimes involving millions of documents
  • Ongoing transition away from heavy reliance on PDFs
  • Uncertainty around how AI tools can effectively support accessibility
  • Accessibility challenges in dashboards and reporting tools (e.g., Power BI)
  • Difficulty distinguishing archived versus active documents
  • Limited evidence on effective AI use cases for accessibility
  • Need for clearer communication of accessibility requirements and stronger internal capacity
  • Accessibility gaps across platforms such as SharePoint, dashboards, and reporting tools

Resources & Examples

Accessibility

ASAP PDF

An AI-powered application for navigating PDF Accessibility audits.

  • Code for America
  • 2025

What accessibility work in your agency is repetitive, text-heavy, and high volume?

  • Managing retention schedules and publishing guidance at scale
  • High-volume PDF remediation
  • Website updates and content conversion
  • Modernizing aging websites to meet regulatory and compliance requirements
  • Managing large volumes of web content in content management systems
  • Preparing external presentations and stakeholder materials
  • Maintaining PDFs, forms, and web applications
  • Converting content to plain language, often requiring human review
  • Making forms and policy documents accessible
  • Creating and updating web content, including plain language and alt text

Resources & Examples

Automation + AI

NIST AI Risk Management Framework Playbook

The AI RMF Playbook offers organizations detailed, voluntary guidance for implementing the NIST AI Risk Management Framework to map, measure, manage, and govern AI risks effectively.

  • National Institute of Standards and Technology (NIST)
  • 2024

What AI accessibility use case are you currently experimenting with?

  • Converting PDFs into accessible web forms
  • Identifying accessibility issues in web pages
  • Supporting PDF remediation with AI tools
  • Using agents for structured web content
  • Developing prompts to rewrite content to align with style guides
  • Evaluating AI models to support document and form accessibility, followed by manual testing

Resources & Examples

Automation + AI

Artificial Intelligence in Public Services

This guide helps UK public bodies understand how to responsibly procure, develop, and use AI while meeting their legal duties to prevent discrimination and promote equality under the Public Sector Equality Duty (PSED).

  • Equality and Human Rights Commission (EHRC)
  • 2022

Where do you worry that AI could create false confidence?

  • PDF remediation that appears complete but misses critical issues
  • Inaccurate or incomplete captions and audio descriptions
  • Automated video captions that do not meet accessibility standards
  • Incorrect or inappropriate AI-generated code and alt text
  • Hidden accessibility issues introduced through AI-supported development
  • Misalignment between AI-generated plain language and established readability standards

Resources & Examples

Automation + AI

GovAI Coalition Templates and Resources

A publicly-available suite of policy templates and knowledge-sharing tools offered by the GovAI Coalition (via the City of San José) to help public agencies launch or refine responsible AI governance programs.

  • City of San Jose
  • 2025

What kinds of tasks would you NOT want automated with AI?

  • Accessibility workflows that require human review and judgment
  • Final decision-making in accessibility implementation
  • Customer-facing service interactions, such as online help desks

Resources & Examples

Are there places where leadership is encouraging AI faster than teams are comfortable with?

  • Limited engagement or acknowledgment of AI in some organizations
  • External pressure influencing AI exploration and adoption
  • Cautious and transparent approaches with mixed perceptions of pace
  • Interest in AI accompanied by ongoing learning and readiness-building

Resources & Examples

What questions do you still have?

  • Impact of agentic, dynamically generated design systems on accessibility
  • Approaches for designing accessible AI agents and assistants
  • Support for ASL and other communication needs in AI tools
  • Applicability of accessibility rules to email
  • Effective models for interagency collaboration on accessibility
  • Impact of varying delivery models and maturity levels on equitable access
  • Addressing gaps between interest in AI and access to implementation
  • Need for plain-language resources explaining AI capabilities and limitations