Remote Work

Remote work is simultaneously an employee preference (consistently high in survey data), a talent strategy (remote-friendly employers attract from a wider pool), a cost structure (real estate and benefits), and a cultural decision. The tension between these forces produces the policy whiplash that has defined the post-pandemic period. Organizations that have thought clearly about which dimension matters most — for them specifically — have arrived at sustainable policies. Organizations that have reacted to each new data point or executive preference have produced policy instability that damages employee trust as much as the underlying policy.

Take a 1,200-person software company operating remote-first since 2020. The operational stack: async-default communication (documented decisions over meetings when possible); quarterly in-person offsites for each team; deliberate timezone distribution (engineering across three major zones); onboarding that assumes no physical co-location; performance management calibrated to output rather than observation; compensation tied to candidate location with transparent geographic bands; technology infrastructure that assumes remote access (cloud-first, SaaS-first, zero-trust security). Contrast: a company that moved to remote in 2020 without changing any of the above and now struggles with culture, onboarding, and performance management — the tooling moved, the operating model didn't.

The single biggest mistake organizations make with remote work is treating it as a policy rather than an operating model change. The policy ("you can work from home 3 days a week") changes nothing fundamental; the operating model ("we run async-default, documented-first, asynchronous-meeting-agenda") is a full transition. Organizations that made the operating model change produced high-performing remote cultures. Organizations that only changed the policy produced the worst of both worlds — employees dispersed, management practices unchanged, and both in-person and remote experiences degraded. The return-to-office backlash is often against a policy framed as remote, not against a real remote operating model.

In knowledge work (tech, services, finance), remote is broadly viable and depends on operating-model discipline.

In manufacturing, frontline roles cannot be remote; corporate functions can, and the operational culture bifurcates between plants and HQ.

In healthcare, direct clinical roles cannot be remote; specific clinical tasks (telehealth, remote radiology, utilization review) are remote-native.

In retail and hospitality, the core workforce is in-location; corporate functions have the same remote/hybrid calculus as any knowledge-work organization.

In public sector, remote has been contested and often constrained by legislative or executive directives requiring in-person presence.

AI reshapes remote work in 2026 by compressing many of the costs that made remote harder. Meeting transcription and summarization mean that async consumers get more context; agent- drafted status updates mean that presence theater becomes less necessary; onboarding agents make the new-hire experience in remote contexts much richer. The remote/in-office productivity debate shifts — the productivity gap, where it existed, was often about coordination overhead that AI substantially addresses. The operating-model question remains, but the operational friction falls.

• Policy without operating model. Letting people work from home without changing how work gets coordinated produces degraded outcomes in both settings.
• Hybrid without clarity. "Some days here, some days there" without coordination turns every day into the worst of both.
• Meeting-first culture in remote. Calendar- filled days in a remote setting are uniquely painful. Async-default is the discipline that works.
• Compensation inconsistency. Location- based bands announced inconsistently produce resentment and attrition.
• Onboarding for co-located assumed knowledge. New hires who never step in an office miss ambient context that co-located onboarding transferred passively. Build the transfer explicitly.