What 2020 proved was not that offices are useless. It proved something more specific, and more unsettling: a vast amount of white-collar output can be produced outside the office.
That distinction matters.
The spreadsheet can be finished at home. The analysis can be written at home. The deck can be designed at home. The recommendation can be made at home. The campaign, memo, forecast, model, presentation, proposal, schedule, report, and design can all arrive as finished deliverables without the employer seeing where the worker sat, what time the worker started, how many interruptions occurred, or what sequence of thought produced the final file.
For a brief period beginning in 2020, millions of employers were forced to confront an uncomfortable experiment. What if the employee could become partially opaque? What if the company could still define the input and judge the output, but could no longer easily observe the hidden process in between?
That is the deeper meaning of the return-to-office debate.
It is not merely a debate about productivity. It is a debate about access.
The Wrong Question
Most RTO arguments begin with the wrong question: Is remote work better or worse than office work?
That framing is too crude. Office presence can matter. Mentorship matters. Culture matters. Apprenticeship matters. Informal learning matters. The hallway conversation is not imaginary. The young employee who learns by watching a senior person handle pressure is experiencing something real.
A serious argument should not pretend otherwise.
But that is not the same as saying the office is necessary for every kind of white-collar output. The better question is not whether offices matter. The better question is: What exactly does the employer lose when the work can be delivered without physical presence?
The answer is not always output.
The answer is often visibility into the mechanism that produces the output.
The Recent RTO Wave
Recent return-to-office announcements make this pattern visible.
In April 2026, Fidelity Investments told Boston employees they would return to the office five days a week beginning in September. Fidelity said being physically together creates more opportunities for connection, mentorship, and learning. Boston.com reported that the move applies to about 6,200 Boston employees, with New Hampshire, Kentucky, and New Mexico offices also affected. Customer-support phone roles were reported as exempt. (Boston.com)
Amazon made a similar move in September 2024, telling corporate employees to return to the office five days a week beginning January 2025. CEO Andy Jassy’s rationale was not that employees had stopped producing files. The stated case was culture, collaboration, brainstorming, invention, learning, modeling, and speed. (CNBC)
JPMorgan Chase directed most employees on hybrid schedules to return to the office five days a week starting in March 2025. The bank’s leadership emphasized learning, innovation, and culture. (Reuters) Dell tightened expectations for its global sales team in September 2024, requiring employees who could work from company offices to do so five days a week, with the stated goal of leveraging a collaborative environment and growing skills. (Reuters) Starbucks shifted support-center partners from three to four days in office and framed the move around reestablishing an in-office culture. (About Starbucks)
Notice the pattern.
The stated rationale is rarely: “The work cannot be done remotely.”
The stated rationale is usually: “The organization needs more access to the conditions under which the work is produced.”
That is a very different claim.
Hybrid Did Not Disappear
The headlines make it feel as though the labor market has simply snapped back to 2019. It has not.
Gallup’s remote-capable-job tracker shows that hybrid work remains the dominant arrangement for remote-capable employees: 52% hybrid, 26% exclusively remote, and 21% fully on-site. Gallup also says these location patterns have remained broadly stable since 2022. (Gallup.com)
The Bureau of Labor Statistics tells a similar story from a different angle. In 2024, 33% of employed people spent some time working at home on days they worked, roughly similar to 35% in 2023. Among workers with a bachelor’s degree or higher, half did some work at home on days worked. (Bureau of Labor Statistics)
So the labor market did not discover in 2024, 2025, or 2026 that remote work was impossible. It discovered in 2020 that a large portion of white-collar labor could be remote or hybrid. What happened afterward was not a simple return to reason. It was a negotiation over discretion.
Who controls where the work happens?
Who controls when the work happens?
Who controls the rhythm?
Who gets to see the process?
Who gets access to the worker before the deliverable appears?
That is the real negotiation.
What the Productivity Literature Actually Says
The productivity evidence is not simple enough to support either slogan.
Stanford’s work-from-home research says working from home rose fivefold from 2019 to 2023, with 40% of U.S. employees working remotely at least one day a week. It also finds that fully remote work is associated with about 10% lower productivity than fully in-person work, citing communication, mentoring, culture, and self-motivation as likely issues. But the same research says hybrid work appears to have no negative productivity effect and can improve recruitment and retention. (SIEPR)
A separate Stanford-linked randomized trial of hybrid work found that hybrid work reduced attrition by 35%, improved satisfaction, and had no significant negative effect on performance ratings or promotions. (SIEPR)
The San Francisco Fed reached a similarly careful conclusion at the macro level. Its researchers found little evidence that the shift to remote and hybrid work has either substantially held back or boosted aggregate productivity growth. (Federal Reserve Bank of San Francisco)
This is why the RTO debate cannot be reduced to productivity. The evidence does not say that offices are useless. It also does not say that remote work failed.
The better interpretation is that productivity is not the whole story.
The Employee as a Black Box
In physics, engineering, and computer science, a black box is a system whose internal mechanism is not visible. You can observe the input. You can observe the output. But what happens inside remains hidden.
That is what remote work revealed about a great deal of white-collar employment.
The employer supplies an input: a request, question, assignment, deadline, target, problem, meeting, or strategic objective.
The employee returns an output: a spreadsheet, memo, deck, recommendation, model, campaign, analysis, design, forecast, report, schedule, or decision.
When the employee is remote, the interior becomes more opaque. The employer can still evaluate the output. It can still ask whether the work was useful, accurate, timely, elegant, profitable, persuasive, or complete. But it has less visibility into the inner process.
It cannot easily see the sequence.
It cannot easily see the rhythm.
It cannot easily see whether the employee struggled, coasted, improvised, collaborated, multitasked, delegated, automated, or used AI.
It cannot easily interrupt the mechanism.
It cannot easily reshape the mechanism in real time.
The black box still produces. But the employer no longer has the same degree of access to what happens inside.
That is the wound RTO is trying to heal.
Employment Was Never Only About Output
This is the uncomfortable truth: the employer-employee relationship has never been purely about deliverables.
If it were purely about deliverables, much more work would have been structured as project work, contracts, piecework, agencies, or outcome-based services. Employment is different because employment gives the organization ongoing access to the person.
The company gets the worker’s attention.
It gets their calendar.
It gets their responsiveness.
It gets their availability.
It gets their presence in meetings.
It gets their participation in rituals.
It gets their exposure to managers, norms, pressure, politics, language, incentives, and culture.
It gets the right to revise not only the work but the worker’s method of working.
This is why the office matters so much to management. The office is not just a building. It is an interface into the employee.
It allows the organization to see, shape, correct, synchronize, and culturally imprint the mechanism that produces the work.
Remote work weakened that interface.
RTO is the attempt to restore it.
The Control Question
A University of Pittsburgh study of S&P 500 firms found results consistent with managers using RTO mandates to reassert control over employees and to blame employees as a scapegoat for poor firm performance. The same study did not find evidence that RTO mandates improved firm value or financial performance, and it found significant declines in employee job satisfaction after mandates. (SSRN)
That does not mean every RTO policy is cynical. Some are sincere. Some are strategically defensible. Some jobs really do benefit from physical proximity. Some teams decay when they never gather. Some younger workers need apprenticeship more than autonomy. Some cultures cannot be maintained by calendar invites and Slack threads alone.
But the control question cannot be dismissed.
When companies say “culture,” they may mean belonging, mentorship, shared standards, and trust.
They may also mean conformity, visibility, surveillance, managerial reach, and reduced employee discretion.
Both can be true.
That is why RTO is emotionally charged. Employees hear “culture” and often suspect “control.” Employers say “collaboration” and often mean “I need access to the process before the output arrives.”
The disagreement is not only about location. It is about authority over the hidden layers of work.
The File-Shaped Future
The deeper reason this matters now is that more work is becoming file-shaped.
A growing share of white-collar value arrives as a digital object: a document, spreadsheet, report, deck, design, model, workflow, codebase, dashboard, proposal, forecast, script, image, campaign, or recommendation.
Once work becomes file-shaped, the worker becomes easier to abstract.
The question shifts from “Where did the person sit?” to “Who or what can produce the deliverable?”
That is where AI enters the story.
AI is not simply another productivity tool sitting beside the worker. AI increasingly occupies the same black-box position that remote work exposed. A request goes in. An output comes out. The internal sequence is difficult for most managers to inspect. The value is judged at the boundary: Was the deliverable useful? Did it solve the problem? Did it arrive on time? Was it good enough to act on?
This is why RTO and AI are colliding at the same historical moment.
Employers are trying to pull human workers back into visibility at the exact moment AI is making more work available as outcome-based production.
That creates an awkward contradiction.
On one side, companies say they need people physically present because culture, mentorship, collaboration, and learning require access to the person.
On the other side, the economics of AI push work in the opposite direction: toward outputs, deliverables, automation, delegation, and black-box evaluation.
The office asks: Can I see the worker?
AI asks: Do you still need to?
The Real Discovery of 2020
The great discovery of 2020 was not that everyone should work from home.
That is too simplistic.
The real discovery was that many organizations had confused three different things:
The work.
The worker.
Access to the worker.
Remote work separated them.
For many white-collar roles, the work continued. The files arrived. The meetings happened. The dashboards updated. The decks were built. The clients were served. The code shipped. The forecasts were made. The recommendations appeared.
But access to the worker changed.
Managers could no longer assume that work was happening inside the old visible container. They could no longer rely on presence as a proxy for attention. They could no longer use office life as the default mechanism for cultural absorption, informal correction, or managerial intervention.
The output survived.
The interface weakened.
That is why RTO became inevitable for some organizations. Not because remote work proved impossible, but because remote work proved too possible.
It showed that output could be separated from presence.
And once that separation became visible, the employer had to decide what it was really buying.
The Office After the Black Box
The office will not disappear. That is not the point.
But the office will have to justify itself more honestly.
It can no longer pretend that every commute is necessary because every deliverable requires physical presence. That argument died in 2020.
The stronger argument for the office is that some forms of human development, organizational trust, apprenticeship, creative collision, and cultural transmission require shared space. That is a serious argument.
But it is not an output argument.
It is an access argument.
It says: we need access to one another before the deliverable exists.
That may be true. But if it is true, employers should say so clearly. They should stop pretending that RTO is simply about productivity. It is about the process through which productivity is produced.
It is about the inside of the box.
The New Divide
The future of work will not be divided cleanly between remote and office.
It will be divided between work whose value depends on visible human process and work whose value can be judged at the boundary of the deliverable.
Some work will need presence because the process is the product: apprenticeship, leadership, hospitality, care, complex negotiation, physical operations, sensitive collaboration, and high-trust creative work.
Other work will continue moving toward black-box production: analysis, reporting, drafting, forecasting, modeling, scheduling, design, campaign generation, research synthesis, documentation, and many forms of administrative coordination.
That second category is where remote work first disrupted the office.
It is also where AI Talent will now disrupt employment.
Once the deliverable becomes the unit of value, the question becomes unavoidable:
Do we need a person in a chair?
Or do we need the outcome?
RTO is one answer to that question.
AI is another.
The next decade of work will be shaped by the collision between those two answers.