The corporate case against remote work has settled, four years on, around a single argument: innovation requires proximity. CEOs from Jamie Dimon at JPMorgan to Andy Jassy at Amazon have invoked the claim to justify five-day office mandates. The evidence base for that claim is unusually thin, and what data does exist is more interesting than the bumper-sticker version suggests. Innovation in modern organizations does require something — sustained attention, intellectual friction, trust dense enough to survive disagreement — but those are properties of how teams are organized, not of where the desks are located. Conflating the two is doing real economic damage.
The argument here: the productivity literature on hybrid work is now substantial enough to draw conclusions, the relationship between collocation and innovation is more conditional than the return-to-office case admits, and the firms that win the next decade will be the ones that figure out how to engineer what offices used to deliver by accident — high-density informal conversation among people who trust each other — rather than the firms that simply mandate desk attendance.
What the empirical record actually shows
Nicholas Bloom's WFH Research team at Stanford has produced the most comprehensive data on the productivity consequences of remote and hybrid work. Bloom's 2022 NBER paper "Hybrid Working from Home Improves Retention Without Damaging Performance" reported the results of a randomized trial at Trip.com (a 1,612-employee experiment): hybrid workers showed no productivity drop and 35% lower attrition than fully in-office workers. A 2024 update extended the finding across more industries. The productivity penalty most CEOs assert remote workers carry is not detectable in randomized data.
The innovation-specific data is more mixed but also more nuanced. Microsoft's 2021 study published in Nature Human Behaviour — based on internal communication data from 61,000 employees — found that fully remote work shifted communication toward stronger ties at the expense of weaker ones, and that fully remote employees had more siloed networks. That is genuine. But the same study did not test hybrid arrangements, and the "25% drop in collaboration" figure widely circulated in the press refers to communication patterns, not innovation output.
Independent research by Carl Benedikt Frey at Oxford's Future of Work programme, and a 2023 paper by Linus Dahlander and colleagues in Research Policy, finds that proximity matters most for the early-stage idea-generation phase of innovation and matters less for execution. That maps onto a hybrid model where teams converge for ideation sprints and disperse for deep work — which is what most high-performing organizations actually do in 2026, regardless of their official "policy."
Why the case for return-to-office gets the mechanism wrong
Proximity is not what offices actually deliver
Cal Newport's A World Without Email (Portfolio, 2021) and Deep Work (Grand Central, 2016) make a different observation about pre-pandemic offices: most knowledge workers in open-plan offices spent enormous fractions of their day in reactive shallow work — Slack pings, meeting hops, email triage. The office did deliver high-bandwidth informal conversation, but in many organizations that signal was already drowned in noise long before COVID. Mandating a return to that configuration without redesigning how attention works is bringing back the wrong half of the office.
Trust, not collocation, predicts innovation
Amy Edmondson's psychological-safety research at Harvard Business School, including The Fearless Organization (Wiley, 2018), consistently finds that the variable that best predicts team innovation is whether team members feel safe to voice ideas, disagree, and make mistakes. The Google Aristotle project — Google's large-scale internal study of what makes teams effective — reached the same conclusion. Psychological safety can be built remotely or in person; it can be destroyed in either configuration. Geography is not the variable.
The "great innovations are slowing" claim has competition from a more boring explanation
The pop hypothesis that big-breakthrough innovation has slowed gets serious empirical support from a different direction. Nicholas Bloom, Charles Jones, and colleagues published "Are Ideas Getting Harder to Find?" in the American Economic Review (2020), documenting that R&D productivity per dollar has fallen significantly across multiple sectors over decades. The slowdown predates remote work by 50 years. Blaming Zoom for what economists were already calling the "Great Stagnation" before the pandemic is misattribution at scale.
What firms that want innovation should actually do
Three concrete patterns show up in the firms that have outperformed on innovation metrics in the post-2020 environment:
Engineer the informal conversation deliberately. Set up regular small-group lunches across teams. Fund weekly off-sites for cross-functional working groups. The "watercooler" was always a thinly disguised forced-contact mechanism; replicating it on purpose works better than waiting for it to happen by accident.
Protect deep work as ferociously as you protect collaboration. Most ideas come from sustained individual thinking that has been informed by prior conversation. The hybrid firms that win in 2026 are the ones that recognize meeting density at 70% of workdays — common in many "in-office" companies — destroys the cognitive conditions for novel ideas as effectively as fragmented Zoom calls.
Build psychological safety as a measurable goal. Edmondson's research provides specific manager behaviors that can be coached and measured. The firms that take it seriously do not treat "team culture" as soft; they treat it as a leading indicator of innovation throughput.
For NWLB's broader argument on what hybrid work looks like five years in — and where the remote-work debate has actually settled — see Remote Work, Year Six →.
Offices used to deliver innovation by accident, through forced-contact density. Mandating their return without rebuilding the underlying mechanism is performance management dressed up as strategy.
"Innovation requires intimacy" is the kind of sentence that sounds profound and means almost nothing as written. Translate it into engineering terms — innovation requires sustained attention, intellectual friction, and trust — and the prescriptions get usefully specific. None of those properties are produced by the act of swiping into a building. They are produced by deliberate organizational design, by managers who protect deep work and convene the right people at the right cadence, and by the kind of cross-functional trust that has to be built whether your team meets in a conference room or in a Zoom grid. The work is doable. It is just not the work most return-to-office mandates are actually asking for.
Updated May 21, 2026. This piece was substantively rewritten as part of NWLB's 2026 editorial refresh.
