The pandemic did not produce one permanent work model. It expanded the set of jobs and organizations that could operate remotely, exposed limits in management and infrastructure, and made work location a continuing design choice.

Employers should avoid treating “remote,” “hybrid,” and “office” as cultural identities. Each is an operating arrangement with effects on access, coordination, learning, security, cost, and worker experience. The right comparison uses a defined job and outcome, not a universal productivity claim.

Read remote-work statistics carefully

The US Bureau of Labor Statistics reported in its 2025 American Time Use Survey summary that 35 percent of employed people did some or all of their work at home on days they worked. That statistic describes activity on worked days; it is not the share of fully remote jobs or a measure of employer policy.

Use precise definitions in internal analysis:

  • fully remote: the employee normally has no required workplace days;
  • hybrid: workplace attendance follows a recurring or role-dependent pattern;
  • occasional home work: some tasks are completed at home without a remote job arrangement;
  • distributed: a team works across locations, which may include offices;
  • home-based platform or piece work: a distinct employment context that should not be merged with salaried telework.

The International Labour Organization’s report Working from Home: From Invisibility to Decent Work covers telework as well as other forms of home work. Its wider lens matters because flexibility, pay, safety, social protection, and worker voice can differ sharply across those groups.

Design around the job and service

Map the tasks that require physical equipment, in-person interaction, secure environments, synchronous coordination, deep individual work, customer presence, or supervised training. Decide location at task level before rolling it into a job-level policy.

Then define the service outcome. A recruiting team might measure time to an accepted interview, candidate access by geography, handoff errors, and recruiter workload. A software team might measure release quality, incident response, onboarding time, and knowledge recovery. Attendance is an input, not the outcome.

Avoid a policy that grants flexibility only to people whose managers negotiate well. Publish eligibility, exceptions, time-zone expectations, equipment, travel, expense, accessibility, and review paths. Record why a role is constrained and revisit the decision when work changes.

Measure collaboration without surveillance

Digital exhaust such as messages, keystrokes, presence indicators, and meeting hours can be easy to collect and hard to interpret. Activity does not establish value, effort, or performance.

Use measures tied to team service: completed work, defects, response commitments, customer outcomes, knowledge transfer, workload, and sustainable pace. Combine quantitative data with structured review. Tell workers what is collected, why, who sees it, how long it remains, and how to challenge an error.

For distributed teams, inspect coordination costs directly. Track decisions delayed by missing owners, repeated meetings, time-zone handoffs, inaccessible documents, and work reopened because context was lost. These signals lead to process improvements without pretending that a green status light measures contribution.

Onboarding and learning require deliberate capacity

Remote work can widen the talent pool while weakening incidental learning if an organization relied on proximity rather than explicit teaching. New employees need role expectations, access, a map of people and systems, sample work, decision records, scheduled feedback, and protected time with experienced colleagues.

Measure time to perform representative tasks, manager and peer load, early errors, access delays, and retention. Compare cohorts carefully: role mix, labor market, manager, and hiring standards may differ.

Hybrid schedules should coordinate the work that benefits from being together. Requiring arbitrary days while teams sit in separate video calls creates travel without collaboration. Plan specific reviews, training, customer work, or relationship-building and test whether the benefit occurs.

Hiring someone in another country can involve employment status, tax, payroll, benefits, immigration, working time, data transfer, intellectual property, and local termination rules. A platform or employer-of-record service can execute parts of the process; the employer still needs to understand responsibilities and employee recourse.

Maintain a location register covering employer entity, contract, payroll provider, manager, work location, permitted data flows, equipment, security, policy variations, and emergency contacts. Do not let an applicant’s willingness to work a time zone substitute for an assessment of sustainable hours.

Global recruiting also changes evidence interpretation. Titles, credentials, salary history, and educational paths differ. Use job-related work evidence and give candidates a way to clarify unfamiliar records.

AI changes tasks unevenly

The ILO’s Generative AI and Jobs: A 2025 Update evaluates exposure at the task level across occupations. Exposure is not a prediction that a job will disappear. Technical possibility, adoption, organizational design, regulation, cost, and worker response all affect what changes.

Inventory tasks before buying broad automation. Classify assistance, recommendation, and execution separately. Record the source evidence and accountable person when AI output affects hiring, evaluation, pay, scheduling, or access to work.

Train people not only to prompt but also to verify, handle sensitive data, recognize limitations, escalate failure, and work without the tool. Measure rework and review burden alongside time saved.

The NIST AI Risk Management Framework is a voluntary structure for governing, mapping, measuring, and managing AI risk. It can organize workforce deployments but does not resolve labor, privacy, or employment obligations.

Run bounded policy experiments

Before changing a work model, record the baseline by role and team: output, quality, customer experience, hiring reach, onboarding, promotion, attrition, absence, employee-reported focus, coordination load, space cost, travel, and security incidents.

Set a defined trial period and comparison. Preserve accessibility and individual exceptions. Avoid forcing teams into conditions known to fail merely to produce an experiment. Include workers and managers in interpreting results.

Document confounders such as restructuring, new leadership, demand shifts, or tool changes. Publish what remains unknown. A result from one function should not become a company-wide rule without checking whether task and workforce conditions match.

A workforce operating scorecard

DimensionEvidence
Accessapplicant geography, accommodation, equipment and connectivity readiness
Servicecompletion time, quality, customer outcome, backlog
Coordinationhandoff delay, meeting load, decision latency, reopened work
Developmentonboarding milestones, feedback, learning access, internal mobility
Sustainabilityhours, workload, absence, retention, worker feedback
Controldata access, security incidents, policy exceptions, correction paths

The post-pandemic workforce is not settled. It is more observable and more configurable than before. Employers can use that opportunity to design around actual work, test claims against outcomes, and make location, skills, and AI decisions that workers can understand and challenge.

Sources and limits

This article uses public data and guidance from BLS, ILO, and NIST. BLS statistics describe the United States, while ILO material spans distinct forms of home work globally. Neither determines the best arrangement for a specific employer or role.