Autistic burnout and the cost of masking, for knowledge workers (not the HR version)

Autistic burnout is not a personal coping failure. It is a predictable output of sustained masking under a work rhythm built for a neurotypical default, and the interventions corporate wellness keeps prescribing for it are contraindicated by the mechanism that produces it. The most common framing — that an autistic knowledge worker has simply hit their “resilience ceiling” and needs to meditate harder, journal more, or “bring their whole self to work” — taxes the exact resource the burnout has already depleted. If you want to stop the recurrence, you have to stop treating the exhaustion as the problem and start treating the mismatch as the problem.

This is the version of the argument aimed at the person quietly losing executive function at their desk, and at the manager trying to read the warning signs without forcing anyone to disclose. It is deliberately not the HR version.

What autistic burnout actually is

The construct most autistic adults and the clinicians who listen to them now use comes from Raymaker and colleagues at AASPIRE (Academic Autism Spectrum Partnership in Research and Education). In the 2020 participatory study — “All I feel is overwhelming exhaustion” — Raymaker et al. defined autistic burnout through three converging features: chronic exhaustion, loss of skills, and increased tolerance to stimulus (meaning, counterintuitively, a reduced tolerance, a shrinking window for sensory and social input before overwhelm). It is a state, not an event, and it recurs because the conditions that produce it usually do not change when the person takes a week off.

The piece generic occupational-burnout frameworks miss is the loss of skills criterion. Maslach’s burnout inventory and its descendants — the frameworks your company’s wellness vendor is almost certainly working from — capture exhaustion, cynicism, and reduced professional efficacy. They do not capture the autistic adult who, mid-burnout, loses access to language they had last quarter, can no longer hold the conversation they used to manage, or finds a script that ran smoothly for years suddenly unreachable. That is not “reduced professional efficacy” in the motivational sense. It is a transient but real loss of capability that recovers slowly, and only when the load actually drops. When an autistic engineer says “I can’t think the way I could two months ago,” the generic burnout rubric has no slot for it, and the resilience-training response has no lever for it.

Raymaker’s team was explicit that autistic burnout is distinct from occupational burnout, and that conflating them is part of why it recurs: you treat the surface (tiredness, disengagement) while the mechanism (sustained masking plus sensory overload plus skill loss) keeps running underneath. The participatory framing matters here — this construct was built with autistic adults describing their own experience, not applied to them from outside. That is the difference between a definition that maps onto the body and one that maps onto a survey.

Masking as a metabolic and cognitive cost

The second load-bearing construct is masking, or camouflaging: the work of suppressing autistic traits and manufacturing neurotypical-coded ones in real time. It is not a personality choice. It is labor, and the cost literature now treats it that way.

Hull and colleagues’ 2017 study, “Putting on my best normal”, documented camouflaging as a conscious, effortful process with a consistent payoff structure: it buys social acceptance and reduces obvious friction, at the price of exhaustion, anxiety, and a corroded sense of self. The Cambridge line of work pushed the cost into harder territory. Cassidy, Bradley, Shaw, and Baron-Cohen’s 2018 finding — that camouflaging was associated with suicidality in autistic adults, and that the association held after controlling for depression — turned masking from an inconvenience into a documented harm. Cage and Troiani, and Monfries, have continued to map the relationship between camouflaging and mental-health outcomes, treating it as a measurable expense rather than a style.

If you read that literature as a knowledge worker, the takeaway is not “masking is bad for you, so stop.” Almost no one can simply stop. The takeaway is that masking has a budget, the budget is smaller and refills more slowly than neurotypical colleagues assume, and a great deal of standard knowledge-work practice is structured to spend that budget on things that produce no output.

The cost of a “good standup” is a useful unit. A standup looks like ten minutes. For an autistic engineer it is ten minutes of real-time performance: monitoring one’s face, calibrating tone, deciding whether to disclose a blocker (will it read as competence or complaint?), manufacturing the right amount of eye contact for the camera, and suppressing the stims that would actually help regulate. That is masking work, and it is paid out of the same pool the rest of the day’s executive function comes from. Multiply by every sync call, every “quick chat,” every hallway interaction in an open-plan office, and the tax is not marginal. It is structural. By 3 p.m. the pool is empty, and the work that needed it — the deep, solitary, actually-valuable work — has nothing left to draw on.

The accelerants specific to knowledge work

Several features of contemporary knowledge work are not merely suboptimal for autistic people; they are precision-engineered to drain the masking budget fast.

Context-switching. Task-switching has a documented cognitive cost for everyone, but for an autistic brain that has built its efficiency around sustained, single-threaded attention, the cost is steeper and the recovery between switches is longer. A calendar carved into 30-minute slots with three different contexts is, for a neurotypical colleague, an annoying day. For an autistic engineer it can be a day in which no actual work happens, because the switching tax consumed the working memory before the task did.

Open-plan sensory load. The evidence on sensory processing in autism is clear that this is not preference but physiology — differences in how the nervous system weights and habituates to input. Open plan is a continuous stream of uncontrolled auditory, visual, and movement signal that an autistic person must actively filter rather than passively tune out. That filtering is masking-adjacent labor, and it runs all day, in parallel with everything else.

Performative visibility. The culture of visible busyness — presence on Slack, reactions in threads, the documented standup, the always-on camera — converts work into a continuous performance of work. Performance is masking. A culture that rewards the signal of engagement over the substance of output systematically overpays the people for whom performance is cheap and penalizes the people for whom it is the expensive part.

Reorg and roadmap uncertainty. Uncertainty about role, scope, and reporting line is a sustained stressor for anyone. For an autistic person whose capacity often depends on a predictable structure, it is a background drain that does not announce itself — it just lowers the ceiling on everything else until the structure is restored.

On-call and incident response. Interrupt-driven, high-stakes, socially dense, and unpredictable in timing. For an autistic engineer it concentrates several of the worst drains into one rotation. Treated as a uniform shared duty, it is one of the more reliable ways to push a capable autistic contributor into burnout.

Why the standard interventions deepen it

Here is where the corporate wellness response fails the mechanism.

Resilience training and mindfulness programs assume the bottleneck is the worker’s internal capacity to absorb load, and that capacity can be expanded with practice. For autistic burnout the bottleneck is external load exceeding a finite masking and sensory budget; “expanding” internal capacity reads, accurately, as a request to mask harder for longer. The intervention spends the resource it claims to protect.

“Bring your whole self to work” campaigns are the more insidious version. They invite the autistic employee to drop the mask — into an environment that is still coded for the neurotypical default, still penalizes the traits the mask was suppressing, and still offers no structural accommodations for sensory load or pacing. The invitation is to be more visible and more vulnerable in a setting that has not changed to receive it. For some, that is a path to genuine relief. For many, it is a trap: the cost of being unmasked in a hostile environment is higher than the cost of masking, and now the worker has been told the choice to keep masking is their own personal failure to “be authentic.” That is a setup, not a benefit.

The pattern across both is the same: the intervention addresses the worker, not the environment. Raymaker’s whole point was that autistic burnout is a mismatch phenomenon. You do not fix a mismatch by training one side of it to tolerate more.

What actually helps

The things that help are less inspirational and more structural. None of them require a diagnosis to be on the table.

Treat disclosure as a calculus, not a virtue test. Disclosure can unlock real accommodations, but the decision is not “be brave and disclose.” It is a calculation about who, where, and what you are asking for. The disclosure study in this network (Newman et al., ASSETS 2025; n=493 software engineers) found the median disclosure outcome is positive — but the forums people take advice from skew heavily negative, and the negativity bias there is real and costly. The right read is not “disclose always” or “disclose never”; it is that disclosure is less risky than the worst-case forum advice suggests, the safest first move is usually to a trusted manager rather than a public channel, and the most common payoff is understanding and reduced masking labor, not a formal accommodation. See Disclosure in software workplaces, beyond the negativity bias for the full version.

Protect deep work as a job design, not a perk. If an autistic engineer’s value is concentrated in sustained, single-threaded work, then protecting blocks of it is not a comfort; it is the operating condition for the output. That means fewer interrupts, batched sync, async-first defaults, and explicit permission to be unreachable during a focus block. This is the single highest-leverage change a manager can make, and it costs nothing.

Take sensory regulation seriously as a legitimate accommodation. Noise-canceling headphones, a quiet space, control over lighting, remote or hybrid as a default rather than a reward — these are not indulgences. They reduce the continuous filtering tax that eats the masking budget. Treat them as infrastructure. On the broader case for designing the environment rather than the person, the strategy subdomain makes the argument that burnout is an infrastructure and mismatch failure, not an individual resilience failure — Building in the gap and the credentials work frame the same shift in a different domain.

Pace, and schedule recovery that is actually recovery. Raymaker’s participants were clear that a long weekend does not repair autistic burnout, and that the recovery curve is long and conditional on load actually dropping. Pacing — building low-load periods into the week as a structural default, not as a reward earned after breakdown — is closer to the right unit than a vacation. If recovery has to be “earned,” you have already lost.

Decouple performance from output. Reduce the standup to what it needs to communicate, written and async where possible. Make presence on chat a signal, not a metric. Every bit of performative visibility you remove is masking budget returned to actual work.

A manager’s read, without forcing disclosure

The constraint a good manager operates under is that they often cannot ask, and should not. Forcing or fishing for a disclosure shifts risk onto the employee and is, in most jurisdictions, a legal landmine. The good news is that most of what helps does not require a label.

The warning signs, read through the burnout lens rather than the engagement lens: a previously strong contributor’s output quietly degrading while their visible activity stays flat or rises (the masking is holding; the work is not); declining tolerance for meetings, interruptions, or ambiguity that used to be handled; withdrawal from optional social load that was previously managed; skill regression — things the person did competently six months ago now taking longer or going wrong. Any of these in isolation is ordinary. A cluster, sustained, is the signal.

The response is environmental, not diagnostic. Do not ask “is something going on with you.” Ask, and act on, the system: can we cut the meeting load, protect a focus block, move this person off the rotation that is clearly draining them, make the standup async, give them a quieter space. Frame the change as an experiment about how the team works, not an accommodation about who they are. If the person wants to disclose, they will, and they are far more likely to when the environment has already shown it can flex without requiring a justification. If they do not, you have still removed load, which is the only thing that was going to help anyway.

The thing this is really about

The reason the HR version fails is that it asks the autistic knowledge worker to locate the problem inside themselves and then solve it inside themselves. Raymaker’s construct, and the masking-cost literature behind it, say the opposite: the problem is a sustained mismatch between a finite budget and an environment that spends it without counting. The autistic engineer quietly losing executive function is not failing to cope. They are running a calculation that has been correct for a long time and is now returning the only answer it can.

That reframing travels. On the strategy side, burnout as infrastructure failure rather than individual resilience is the same argument at the level of the firm: stop optimizing the worker, start optimizing the system that loads them. On the learn side, the same logic reaches assessment design — an assessment that penalizes masking-by-default, that rewards performative fluency over actual understanding, is baking this exact burnout into the credential before the person ever reaches a job. The mismatch is not a workplace problem or a school problem. It is a design problem, and it is the same design problem everywhere it appears.

The autistic adult in burnout does not need to be told to be more resilient. They need the load to come down. Almost everything that follows, for them and for the people managing them, is a question of whether the environment is willing to change before the person is required to.