How should we rethink accessibility, AI, and human judgment in EdTech?
The short answer: Accessibility in higher education works best when it’s built into the content creation process from the beginning, rather than treated as a compliance checkbox at the end. AI can accelerate the work, but human judgment remains essential for context, intent, and quality.
TL;DR
- Accessibility should be built into content creation from the start—not retrofitted at the end.
- AI can identify accessibility issues faster, but human judgment is required for context and intent.
- Institutions should prioritize high-enrollment courses, assessments, and known accommodation needs instead of remediating everything at once.
- Responsibility for accessibility must be shared across the institution, not siloed to a single team.
- The goal is steady, meaningful progress, not perfection.
This post is based on a Blackboard panel discussion featuring April Akins (Blackboard Ally & Assistant Blackboard Administrator, Greenville Technical College), Amy Lomellini (Director of Accessibility, Blackboard), and Daniel Singletary (Digital Accessibility Manager, University of Missouri System), moderated by Julie Uranis (SVP of Online and Strategic Initiatives, UPCEA).
Watch the full panel discussion here: From Checklists to Check-ins: Accessibility, AI, and Human Judgement in EdTech | Blackboard
What does meaningful accessibility look like in the age of AI?
Meaningful accessibility means building inclusion into the design of learning content from the start—not discovering barriers after the fact. In an AI-enabled environment, it means using automation to catch issues faster while keeping humans in the loop for decisions that require context.
For years, higher education treated accessibility as a compliance exercise. That framing was a starting point, but with digital learning now central to the student experience, it is no longer sufficient.
“You’re trying to build accessibility as your foundation and not as the ceiling to reach.”
—April Akins, Ally & Assistant Blackboard Administrator, Greenville Technical College
Why should accessibility be built in from the start, not retrofitted?
Retrofitting accessible content is slower, more expensive, and less effective than designing it correctly the first time. The same principle applies in higher education as in architecture.
“When accessibility is built in upfront, it provides equal access for everyone and avoids the time, cost, and effort involved in retrofitting materials later.”
—Daniel Singletary, Digital Accessibility Manager, University of Missouri System
When accessibility is left to the final step, issues surface too late, leaving instructors under pressure and students facing barriers. Treating it as an ongoing process means problems are caught earlier and good design becomes the default.
Where does AI help with accessibility, and where does it fall short?
AI is effective at identifying accessibility problems at scale and reducing manual effort. It can generate alt text, flag issues in documents, and convert materials into more accessible formats. But it cannot supply context or intent.
- AI can describe what is in an image. It cannot explain why that image matters in the context of a lesson.
- AI can convert documents. For equations, diagrams, and missing source files, human review is still required.
- AI can flag reading order problems. Determining the correct order for complex layouts requires human judgment.
“When AI is just doing something without the human in the loop, it can create more harm than good.”
—Amy Lomellini, Director of Accessibility, Blackboard
Which courses and materials should institutions prioritize for accessibility remediation?
Institutions that try to remediate everything at once often miss the barriers that matter most. A focused approach produces more meaningful results.
Prioritize in this order:
- Required and high-enrollment courses: Greatest impact on the most students
- Assessments and core activities: Barriers here directly affect grades
- Materials tied to known accommodation needs: Address documented student requirements
- Native, flexible formats over static PDFs: Easier to maintain and update
“Meeting the numbers” does not equal real progress. What matters is whether students actually experience fewer barriers—not whether remediation totals look good on a report.
How should institutions share responsibility for accessibility?
Accessibility often fails when it is owned by a single team. It succeeds when it is distributed, embedded in the everyday work of instructors, instructional designers, and administrators.
Community-driven models—accessibility champions, peer advocates, embedded training—help spread ownership and make improvement continuous. Accessibility is not a project with an end date; more people involved means more sustained progress.
AI supports this by helping teams catch issues earlier and freeing time for better decisions, but technology cannot replace the judgment, context, and lived experience that accessibility depends on. The question should never be whether something is done—but whether it is actually working.
Watch the full panel discussion here: From Checklists to Check-ins: Accessibility, AI, and Human Judgement in EdTech | Blackboard
