The Future of Accessibility Is the Future of Work

Wiberg's story matters here not because it is dramatic, but because it is ordinary in a specific way. He was not hired to build an accessibility feature. He was living with vision loss, understood the actual shape of the problem better than any outside designer could, and built the solution he needed. That distinction, between technology built for disabled people and technology built by them, sits at the center of a debate that has quietly become one of the most consequential in the future of work.

The debate has real stakes because investment in accessibility technology is accelerating almost everywhere at once. Employers are buying AI powered hiring platforms. Technology companies are shipping captioning, screen reading, and voice control features as standard rather than optional. Governments are writing new guidance on how automated decisions intersect with disability rights. What is much harder to find is rigorous evidence that any specific technology actually improves employment outcomes for disabled workers, as opposed to simply feeling like progress. That gap between investment and proof is not a reason to dismiss the effort. It is the honest starting point for understanding where things actually stand.

Start with the plainest measure available. In 2025, 22.8 percent of people with a disability in the United States were employed, compared with 65.2 percent of people without a disability, according to the Bureau of Labor Statistics. That gap was little changed from the year before. The unemployment rate for people with a disability rose to 8.3 percent over the same period, more than double the 4.1 percent rate for people without a disability. These are not soft or interpretive numbers. They come from the government's own monthly survey of American households, and they describe a labor market in which disabled workers remain locked out at roughly three times the rate of everyone else.

It is easy to find a more encouraging statistic sitting right next to that one. It is worth being precise about what it actually measures. Employment among people with disabilities grew by nearly 25 percent from pandemic era lows, reaching more than 7.3 million workers by the middle of the current decade. That is a real gain in the number of disabled people holding jobs. It is a different thing from the employment ratio above, which measures what share of the entire working age disabled population is employed. A rising headcount and a flat ratio can both be true at once, and in this case they are. More disabled people are working than in 2020. The proportion of disabled people able to work who are actually employed has barely moved.

Nowhere is the gap between technological momentum and proven outcome sharper than at the point of hiring itself. In an October 2024 survey of more than 900 business leaders, 83 percent said their company used some form of AI to screen resumes, whether that meant software that scores applications, timed cognitive tests, or video interviews analyzed by machine learning models. Estimates like that vary widely across surveys depending on how AI is defined, but the direction is not in dispute: algorithmic screening has become common at the exact point where an employer first decides who gets considered. In May 2022 the Equal Employment Opportunity Commission and the Department of Justice each issued technical guidance warning that these tools can violate the Americans with Disabilities Act. A system trained on how a typical candidate is expected to type, speak, or move can screen out a qualified disabled candidate without anyone ever deciding to discriminate against them. The discrimination happens inside the model, not in a hiring manager's head.

That warning is no longer theoretical. In Mobley v. Workday, a federal judge in California allowed claims to proceed that treat Workday's applicant screening software as an employer's agent. That means the software vendor itself, not just the companies using it, can be held responsible for discriminatory outcomes. In May 2025 the court certified a nationwide collective action covering claims that the tool disparately affects candidates by disability, race, and age. The case is still working its way through the courts, and it has not yet produced a final ruling on the underlying discrimination claims. But it has already established something important: the question of who is accountable when a hiring algorithm screens out disabled candidates is now a live legal question, not a hypothetical one.

The pandemic produced one of the few employment shifts for disabled workers that shows up clearly in the data, and it did not come from a hiring platform or an accessibility feature. It came from a shift in where and how work happens. The Kessler Foundation's 2022 National Employment and Disability Survey found that the pandemic increased employer use of remote work, flexible scheduling, and job sharing as accommodations for disabled employees. Often this happened because the same arrangements were suddenly being offered to everyone, disabled or not. For workers who had spent years being told that remote work was not possible for their role, watching it become standard practically overnight was its own kind of evidence.

That gain was not distributed evenly, and it is worth resisting the temptation to describe it as a universal win. In May 2020, only 25.7 percent of employees with disabilities were working remotely, compared with 35.8 percent of employees without disabilities. Workers with disabilities did not see their remote work access rise at the same rate as their nondisabled peers as the pandemic went on. The disabled workers least able to take advantage of the shift tended to be men, racial minorities, and people with less formal education. They were often concentrated in physically demanding or direct service roles, where remote work was never going to be an option no matter how flexible an employer became. Remote work helped a specific slice of the disabled workforce a great deal. It did not touch a large part of it at all.

There is a pattern in accessibility history that predates any of this technology, and it helps explain why the distinction between building for disabled people and building with them matters so much. In 1945, Kalamazoo, Michigan became the first American city to cut ramps into its sidewalk curbs, responding to requests from disabled veterans returning from the Second World War. Those curb cuts were designed for wheelchair users. Within a generation they were being used by parents pushing strollers, travelers pulling luggage, and delivery workers pushing hand trucks. Policy analyst Angela Glover Blackwell named this pattern the curb cut effect in a 2017 essay: solutions built for one group end up serving a much wider public. Versions of it show up constantly in technology. Captions built for deaf viewers now serve anyone watching a video without sound. Voice control built for people with limited mobility now sits inside nearly every phone. It is worth naming a real tension in that framing, though: describing accessibility gains mainly through how much they end up helping nondisabled people risks treating disabled people's own needs as insufficient justification on their own. The curb cuts were the right thing to build because disabled veterans asked for them, not because parents with strollers eventually benefited too.

A handful of large employers have tried to build the same underlying principle directly into hiring rather than waiting for it to happen by accident. SAP's Autism at Work program, launched in 2013, now operates in twelve countries, employs roughly 150 autistic people, and reports a 90 percent retention rate, a company-reported figure rather than an independently audited one. Microsoft's Neurodiversity Hiring Program, running since 2015, replaces the standard interview with a multi day, hands on skills assessment for neurodivergent candidates. That is an acknowledgment that a conventional interview format measures comfort with conventional interviews more than it measures ability to do the job. JPMorgan Chase built a similar program the same year. It grew from a four person pilot into partnerships across at least ten lines of business. The company has reported that its first Autism at Work cohort was up to 92 percent more productive than peers on comparable tasks, though it has not published the methodology behind that figure, and it describes one early cohort, not a controlled or independently verified study. None of these programs prove that neurodivergent hiring is a universal solution. What they demonstrate is narrower and more useful: that when the design of a hiring process itself is questioned rather than assumed, different people succeed.

Even employers who genuinely want to know whether their accessibility investments are working run into a basic problem: they often cannot tell who is disabled in their own workforce. In the 2025 Disability Equality Index, a benchmarking survey completed by 655 companies across eight countries, participating employers reported a 29 percent increase since the year before in making sure interview candidates knew they could request accommodations. That is a real and measurable improvement in process. But voluntary self identification of disability status among employees at those same companies stayed at a median of roughly 3.5 to 4 percent internationally, and that number did not move in 2025. Disability prevalence in the general population is far higher than that. Part of that gap is a measurement problem. Part of it may not be. Disclosing a disability at work carries a real, documented risk of stigma and discrimination, and choosing not to disclose can be a rational response to that risk rather than a data collection failure an employer can simply survey its way out of. Either way, the gap between the two figures means that most companies publishing accessibility commitments are doing so without a reliable count of the population those commitments are meant to serve. That is precisely why claims that a given technology or program improved employment outcomes for disabled workers deserve real scrutiny rather than automatic trust. The investment is visible. The proof, in most cases, is not yet.

For employers, the practical lesson is not to abandon algorithmic hiring tools but to stop treating them as neutral. A screening system trained on historical hiring patterns will tend to reproduce whatever biases, including disability bias, shaped those patterns in the first place. The ADA already makes employers responsible for that outcome regardless of which vendor built the tool, a point the EEOC's 2022 guidance made explicit rather than invented. Auditing a hiring pipeline for disparate impact on disabled candidates is no longer optional due diligence, and the Workday litigation is a preview of what happens to employers who treat the audit as optional rather than necessary.

For technology companies, the lesson sits closer to Wiberg's story than to any compliance document. Consider the products that have most reliably improved life for disabled people: captions, screen readers, curb cuts, video relay tools like Be My Eyes. Each was shaped early and directly by people who needed it, not retrofitted afterward by a team studying disability from a distance. Neurodiversity hiring programs at SAP, Microsoft, and JPMorgan Chase work the same way. They change who is in the room designing the process, not just who the process is aimed at. A company that wants its accessibility features to actually work has reason to hire the people those features are meant to serve, not only survey them.

For disabled workers themselves, the clearest takeaway may be the least technological one. Remote and flexible work arrangements expanded fastest during the pandemic not because a new accessibility technology arrived, but because employers were forced to prove that flexibility they had long called impossible was, in fact, possible. That history is worth remembering the next time an employer says a given accommodation cannot be done. It usually can. Whether it will be depends less on the state of the technology than on who has the power to ask for it and be taken seriously.

None of this adds up to a tidy verdict. The honest state of the evidence is this: accessibility investment is accelerating faster than anyone's ability to measure whether it is working. The employment gap between disabled and nondisabled workers has moved only modestly, despite a decade of new tools, new guidance, and new corporate programs. What has moved more clearly is something harder to quantify. A Danish furniture craftsman built the accessibility tool he needed. A handful of hiring programs let their design be shaped by the people who would use them. Together, they point to the same conclusion: accessibility works best when disabled people are treated as its authors, not only its intended audience. Nothing About Us Without Us was a disability rights slogan before it was a design principle. The evidence assembled here suggests it may end up being the more reliable measure of progress than any single technology ever will.

Key Findings

• ✓People with a disability had a 22.8 percent employment rate in 2025, versus 65.2 percent for people without a disability, a gap essentially unchanged year over year (U.S. Bureau of Labor Statistics).
• ✓The EEOC and DOJ warned in 2022 that AI hiring tools can violate the ADA, and Mobley v. Workday is now testing in federal court whether AI vendors themselves can be held liable for discriminatory hiring outcomes.
• ✓Remote work expanded disability accommodations during the pandemic, but access to it remained unequal: disabled workers trailed nondisabled workers in remote work rates even as the option became more common overall.
• ✓Neurodiversity hiring programs at SAP, Microsoft, and JPMorgan Chase report strong retention and productivity results, though these are company-reported figures for small, specific populations, not independently audited or broadly generalizable data.
• ✓Median voluntary self identification of disability status among employees stayed at roughly 3.5 to 4 percent in the 2025 Disability Equality Index and did not improve year over year, limiting how precisely any employer can measure its own progress.

16 statements support this investigation, drawn from 15 sources.

Key Evidence

• Verified·95% confidence·March 3, 2026

  In 2025, 22.8 percent of people with a disability in the United States were employed, compared with 65.2 percent of people without a disability, a gap that was little changed from the prior year. Source

• Verified·95% confidence·March 3, 2026

  The unemployment rate for people with a disability rose to 8.3 percent in 2025, more than double the 4.1 percent rate for people without a disability. Source

• Verified·90% confidence·May 12, 2022

  In May 2022 the EEOC and the Department of Justice each issued technical assistance warning that employers' use of algorithmic and AI hiring tools can violate the Americans with Disabilities Act by screening out qualified candidates with disabilities. Source

• Verified·85% confidence

  In May 2020, only 25.7 percent of U.S. employees with disabilities worked remotely, compared with 35.8 percent of employees without disabilities, and workers with disabilities did not gain remote-work access at the same rate as nondisabled workers after the pandemic began. Source

• Verified·85% confidence·January 1, 2017

  Policy analyst Angela Glover Blackwell named this pattern the curb cut effect in a 2017 essay published through PolicyLink, the racial and economic equity organization she founded, describing how solutions built for one group end up benefiting a much wider public. Source