Who Has to Ask Permission to Work From Home?

Billesdon had worked at Wells Fargo for nearly thirty years when the company sent everyone home in 2020. He had a paralyzed bladder and colon, a condition that made an ordinary office day a logistics problem, and for the first time in his career he didn't have to think about it. He later described working from home as "a big blessing," because he could work next to a bathroom without asking anyone's permission first.

In the summer of 2021, Wells Fargo set a date to bring people back. Around the same time, the manager who had quietly made his arrangement work for a year left the company. Billesdon worried that whatever he had going forward would be informal and fragile, something he'd have to renegotiate every time his symptoms flared. So on August 31, 2021, for the first time in his career, he filed a formal request: permanent accommodation to work from home.

According to the trial record, senior managers reacted to the request with surprise and skepticism. They asked whether accommodations management was new to the company and whether legal was involved. They asked why he hadn't brought the request forward before. They discussed what they would need to "push back" against it. When an accommodations consultant suggested giving him a trial period to test the arrangement, management rejected the idea. His performance rating, despite record sales that year, was quietly downgraded from "exceeds" to "meets."

Then Wells Fargo delayed its return date indefinitely, and told Billesdon his accommodation need was already being met by the company's general work-from-home policy. In January 2022, without ever ruling on his request, the company closed his case. A month later, Wells Fargo announced a firm return-to-office date. Fifteen days after that, Billesdon was fired.

A jury awarded him more than $22 million. On appeal, the outcome split in a way that turns out to matter more than the number did. The Fourth Circuit reversed the failure-to-accommodate verdict, but for a specific and instructive reason: because Billesdon had, in fact, been working from home continuously from the day he asked until the day he was fired, the court reasoned that Wells Fargo had never actually failed to accommodate him. It had just never decided the question. The court's opinion states the underlying principle directly: a benefit doesn't stop counting as an accommodation just because everyone else has access to it too. The retaliation verdict survived, on separate evidence, including how the company had handled the request once the universal policy that had been quietly covering him was about to disappear.

One lawsuit does not prove how often this happens. What it does prove is that the mechanism is real, that it has already reached a federal appeals court, and that the court's own reasoning names it: a company can provide someone with everything an accommodation is supposed to provide, for years, without the question of whether they're entitled to it ever actually being decided, right up until the moment the surrounding policy that made the question unnecessary goes away.

The working hypothesis behind this investigation was straightforward. If remote work had genuinely opened employment to disabled workers during the pandemic, then the return to offices ought to show up as a reversal: employment falling, participation dropping, the gains erased. That would have been a clean story. It is not what the evidence shows.

The employment-population ratio for working-age disabled Americans rose from 30.9 percent in 2019 to 38.1 percent in 2025, a 23 percent proportional increase, against essentially flat growth for non-disabled workers over the same period. The most rigorous causal study available, published in the American Economic Review in June 2026, estimates that the post-pandemic rise in remote work explains between 68 and 85 percent of the increase in full-time employment specifically among people with physical disabilities.

That number needs a qualifier. A separate Federal Reserve analysis found that roughly a third of the overall employment gain reflects more people newly identifying as disabled after 2020, not purely new job opportunities for people who were already disabled, meaning the headline increase is real but partly a measurement effect. Neither finding tells us what happened next. As of the most recent data available, no study has isolated the causal effect of return-to-office mandates on disability employment specifically, in either direction, because the infrastructure to even ask that question doesn't exist: none of the labor-market data everyone currently uses tracks disability status. The researchers who track this monthly describe the current state with a word that matters: a plateau, not a collapse.

That absence should not read as a loose end. It's the reason this investigation changed direction.

If the employment totals held up, then the interesting question isn't whether disabled people lost their jobs. It's whether something else changed underneath a number that looks stable.

There is a real difference between two arrangements that can look identical from the outside. Under the first, remote work is simply how the job is done: nobody has to disclose a disability, produce medical documentation, or explain why an office presents a barrier, because the arrangement is built into how the role works for everyone. Under the second, remote work is an individual accommodation: a disclosure, sometimes medical paperwork, an interactive process with a manager or HR, an analysis of what the job actually requires, and a real chance of delay or denial. The same laptop, the same video calls, the same output. A completely different institutional relationship to the person doing it.

Billesdon's case sits exactly on that seam. What moved him from the first category to the second wasn't a change in his medical condition or his job performance. It was the disappearance of the general policy that had been quietly doing the accommodating for him.

Federal disability law has required employers to consider telework as a reasonable accommodation since at least 2003, when the EEOC issued formal guidance on the subject. Employers don't have to grant it, and they can offer an equally effective alternative instead. That guidance predates the pandemic by nearly two decades, and it was never really tested at scale until 2020 forced the question across vast parts of the economy at once.

The EEOC's own pandemic-era guidance contains a real, unresolved tension about what that mass experiment should count for. One part warns that temporarily excusing an employee from in-person work during an emergency doesn't mean an employer has permanently redefined the job; restoring the prior arrangement is allowed, and any renewed accommodation request starts from the usual rules. Another part says the opposite is also true: that period of remote work "could serve as a trial period" showing whether someone can actually do the job from home, and that a later request should be considered in light of that evidence. Courts deciding recent cases have mostly leaned on the first idea rather than the second. A 2015 federal appeals ruling, which ultimately sided against the disabled employee's own request for remote work because the record before that specific court did not show the job could be done from home, had, in its reasoning, still anticipated the broader argument: it said crediting new technology with changing what a job requires would take an actual record showing the technology works for that specific job, not just an assumption that it might. The pandemic built exactly that record for millions of jobs. Whether it counts as evidence, or as a footnote to be set aside once things go back to normal, is still being argued in court.

None of this shows up as a single dramatic event. It shows up as volume.

A 2025 survey of nearly 350 U.S. employers by the law firm Littler found that among companies that had recently required more days in the office, 56 percent reported an increase in requests for remote-work accommodations, rising to 65 percent among the largest employers. That is not evidence that any particular request was wrongly denied. It's evidence that tightening a universal policy pushes the same underlying need back through the individual, evidentiary channel, exactly as the mechanism in Billesdon's case would predict, and employers are the ones reporting it.

Inside the federal government, that channel visibly clogged. An internal Treasury Department memo, later filed in federal litigation, disclosed a backlog of some 6,500 pending reasonable-accommodation requests, most of them involving the IRS. The Department of Health and Human Services reported a backlog of roughly 9,000 requests. None of this proves every request should have been granted. It does mean thousands of individual disability determinations, previously unnecessary, suddenly needed to happen at once, through a process that wasn't built for that volume.

Two current federal employees are testing exactly this in court right now. According to their complaint, one has Type 1 diabetes and severe migraines and had been teleworking since 2020 under an approved accommodation; the other has Stage IV lung cancer and had her own approved accommodation revoked roughly a year after it was granted. Both had their telework arrangements pulled after a 2025 federal return-to-office directive, and both were denied when they asked to have them reinstated. A federal judge granted a preliminary injunction in July 2026 ordering the government to restore their telework while the underlying lawsuit continues; the case is not yet resolved, and the plaintiffs are seeking to represent a larger group of employees who say the same thing happened to them. Whatever the final ruling, the injunction alone confirms the mechanism is live right now, not a historical curiosity from the first year of return-to-office mandates.

The government's own regulators have started describing this problem in their own words. Guidance issued by the EEOC cautions federal agencies against exactly the pattern in Billesdon's case: simply canceling a previously granted telework arrangement and telling the employee to file a new request from scratch. In the same document, the EEOC acknowledges that many original decisions to grant telework "were made without sufficient information," meaning an agency can lawfully ask for fresh medical documentation precisely because it never rigorously evaluated the arrangement the first time around. That is the mechanism, stated by the regulator that enforces it: the universal policy meant nobody had to build the individual case, and its disappearance means the case has to be built now, after the fact, under less generous circumstances.

Large employers routinely describe themselves as disability inclusive, and many participate in formal benchmarking programs built to prove it. The leading one in the United States, Disability:IN's annual index, has asked companies whether they offer flexible work options every year since 2022. The answer has come back between 96 and 99 percent yes, every single year, including the most recent cycle, which slightly reworded the question to specifically mention remote accommodations for employees with disabilities. A question that returns the same near-universal answer for four consecutive years isn't measuring anything that could tell you whether a company preserved remote roles or mandated five days in office with case-by-case exceptions. The index's own newest report states that a more detailed flexible-work question is planned for its next cycle, which is itself an acknowledgment that the current one doesn't capture much.

Among the other disability-inclusion frameworks this investigation examined, in the UK, Canada, and Australia, none asks a more specific question about remote-work availability, hybrid access, or participation mode, and none measures the commute itself, the physical journey between a worker's home and wherever an employer has decided the job happens. That's worth sitting with. It means an accessibility capability can be quietly withdrawn from an organization without that withdrawal ever registering as a setback in the very systems built to measure whether the organization is accessible. The withdrawal can therefore occur without registering clearly in the inclusion systems this investigation examined, because those systems were not designed to measure participation mode at that level of detail.

None of this is an argument that offices should disappear, and the honest version of this story has to take the case for physical presence seriously rather than waving it off.

Isolation is a real cost of remote work, not a manufactured one. Mentoring happens more easily in person. New employees learn a culture faster by being physically embedded in it. Some work genuinely requires being in a specific place, operating specific equipment, or standing next to a specific person. None of that is in dispute, and nothing in this investigation suggests otherwise.

What the evidence doesn't support is the leap from "physical presence helps many people" to "physical presence should therefore be required of everyone, in every role, regardless of what the work actually demands." The alternative this investigation's evidence points toward isn't universal remote work. It's multiple genuinely available participation modes, wherever the underlying work can actually support more than one, decided by what the job requires rather than by which policy is administratively simplest to enforce company-wide.

There's a design assumption buried under most of this debate: that a job begins at the office door, and that everything before that point is the employee's own problem to solve. For a worker who relies on paratransit, who experiences debilitating fatigue or pain from travel, who is immunocompromised, or who needs personal-care logistics that don't scale to a daily commute, that assumption is doing real, uncounted work. Whether an employee can perform a job's actual functions is a different question from whether they can repeatedly reach the specific location an employer has chosen to place that job in, and current accessibility frameworks, so focused on the building itself, largely treat the second question as if it doesn't exist.

Remote work isn't automatically the fix, either. A remote arrangement built on inaccessible software, poor captioning, a screen reader that doesn't work with the company's intranet, or a video platform that requires a camera on at all times, can recreate exclusion in a different shape rather than eliminating it. The goal implied by this evidence isn't remote participation for its own sake. It's accessible participation, in whichever mode actually works, and neither mode gets there automatically.

Before 2020, whether a given job could genuinely be done remotely was often a matter of guesswork and institutional habit rather than anything anyone had actually tested. The pandemic forced an unplanned, unprecedented experiment across nearly every industry at once, and for millions of roles, it left behind something that didn't exist before: an actual operational record of what happened when the job was done from home. That record doesn't settle whether every job should be remote, or whether any specific person is entitled to work remotely forever. What it changes is the shape of the argument the next time the question comes up. The 2015 appeals court that said crediting new technology required an actual record wasn't wrong. That record was missing then. It exists now, for a lot of jobs, sitting in payroll and performance data most companies already have.

The same shape shows up outside of employment. Roughly one hundred of the largest U.S. school districts have been tracked continuously since March 2020, and the number offering any form of remote learning to all students fell from 56 out of 100 in the 2021-22 school year to 46 out of 100 the following year, with several districts explicitly narrowing eligibility to students who met attendance and academic thresholds first. When California passed a law in 2021 restricting districts to in-person instruction with limited exceptions, fifteen students with disabilities sued, arguing they were being denied an education they had been receiving safely for a year. A federal court agreed and ordered the arrangement reinstated while the case proceeded, on the basis that the districts already knew how to provide it and had simply been told to stop.

The clearest empirical version of this same mechanism, showing up as a study rather than a lawsuit, comes from the United Kingdom rather than the United States: a peer-reviewed study of over three hundred university students, conducted amid reports that some universities have stopped recording lectures after returning to in-person teaching, found that nearly half of all respondents said they relied on that flexibility, a reliance reported across disabled, neurodivergent, and other students alike, forcing the same shift this investigation found in employment, from something built into how the class worked for everyone into something that required registering with a disability office first. The comparable U.S. data doesn't exist in the same form: the federal government's own school-level tracking of remote and hybrid instruction offerings, run continuously since 2021, has never once recorded whether a student receiving that instruction had a disability. That's not a minor omission. It means nobody in a position to notice this pattern at a national level in American schools currently has the data to see it.

The deeper question the education side raises, more clearly than the employment side does, is why access that could reasonably be made available to everyone by default should require a formal disability disclosure before someone is allowed to use it at all.

None of this should be read as a description of a golden age that then ended. Even during the period of universal remote work, access to it was sharply uneven. One peer-reviewed analysis found that disabled and non-disabled workers had similar rates of remote access before the pandemic, but that disabled workers had actually fallen behind non-disabled workers on that measure by 2021, and that whatever access existed within the disabled population was concentrated among workers with four-year degrees, disproportionately white, and disproportionately women. Some disabled workers, in service, manual, and frontline occupations, never had a remote option to lose in the first place. Others found that the physical workplace itself had accommodations, equipment, or support that simply weren't replicable at home. The story here is not "everyone had it and then some people lost it." It's narrower and more specific than that: for a particular, unevenly distributed slice of the disabled workforce, mostly in office-based, credentialed roles, something that had briefly stopped requiring individual justification started requiring it again.

Nothing forces this conclusion directly from a single document or study; it's the synthesis this investigation arrived at after everything above. Institutions appear to have treated two different things as though they were the same thing: remote work as a temporary pandemic operating model, adopted for safety and expected to end, and remote work as accessibility infrastructure, which for some workers had quietly become a durable answer to a real, ongoing barrier. When the first was withdrawn, because the emergency had passed and businesses wanted people back, the second went with it by default, not because anyone decided it should.

The failure, on this reading, wasn't asking people to come back to the office. It was never deciding, while the universal policy still existed, what part of it was worth keeping on purpose, for whom, before the whole thing got switched off at once.

Every recommendation that follows traces back to a specific problem this investigation documented, not a generic flexibility wish list, and none of them describes an existing legal requirement. Current law requires employers to consider telework as a possible accommodation; it does not require them to treat a prior remote-work track record as binding evidence in a later request, which is why that idea appears here as something the evidence suggests institutions should consider, not as a rule they are already bound by.

Where a job's actual functions can be performed remotely and an employer already has years of evidence that they can, that history could reasonably count for something the next time the question is asked, rather than being treated as if it never happened, the exact gap the EEOC's own 2026 guidance leaves open today. Federal and large private employers facing backlogs in the thousands need enough staffing to process requests within the timelines they've already set for themselves. Where genuine uncertainty exists about whether a remote arrangement will work, a trial period, the same proposal Wells Fargo's own accommodations consultant made and management rejected, is a lower-cost way to find out than an outright denial followed by years of litigation. Disability-inclusion frameworks that already ask detailed questions about representation, recruitment, and supplier diversity could ask an equally specific one about participation modality and commuting access, rather than a single saturated yes-or-no question that hasn't changed in four years. And in education, where a recording or a remote option can reasonably be made available to an entire class without meaningfully changing what's being taught, requiring a formal disability disclosure before a student can use it adds a barrier that didn't need to exist.

Billesdon had already been working from home, successfully, for a year and a half, by the time anyone at Wells Fargo formally considered whether he should be allowed to. The company had, without meaning to, already run the experiment and already had the answer. What the case shows is that having the answer and having to prove it are not the same thing, and that the second can reappear at any moment the first stops being universal.

The question this investigation keeps landing on isn't whether everyone should work from home. It's what an institution owes to a capability it has already demonstrated, in its own operations, that it knows how to provide, once it stops being something everybody gets and becomes something one person has to ask for again.

Recommended Actions

• ✓Employers with a multi-year remote-work track record for a role should consider treating that history as relevant evidence in future accommodation decisions, as a practice, not because current law requires it.
• ✓Federal and large private employers facing accommodation backlogs in the thousands should staff the review process to meet the processing timelines they have already committed to.
• ✓Disability-inclusion benchmarking organizations should consider adding a decomposed question on remote-work and hybrid-access policy rather than relying on a single saturated yes-or-no measure.
• ✓Where genuine uncertainty exists about whether a remote arrangement will work, offering a trial period is a lower-cost way to find out than an outright denial followed by litigation.
• ✓Educational institutions should evaluate whether recordings or remote-attendance options that can reasonably be made available to an entire class should require individual disability disclosure at all.