The return-to-office debate has already been decided. Most people just haven’t read the verdict.
The headlines still suggest a tug-of-war. Mandate announcements at JPMorgan, Amazon, Disney. Politicians using “back to the office” as a soundbite. Founders posting LinkedIn screeds about productivity. The noise implies the question is still open.
It isn’t. The numbers settled it in 2025.
The Numbers Already Decided This
Gallup’s latest tracker puts 52% of remote-capable US workers in hybrid arrangements, 27% fully remote, and just 21% fully on-site. 25% of all paid US workdays are now performed from home — up from under 5% pre-pandemic, and statistically flat for two years. FlexJobs reports remote job postings up 20% in Q1 2026 quarter-on-quarter, with growth strongest in mid-career roles. The talent market is repricing time, location, and presence as separate variables — not bundled into a single contract called “the job.”
A floor has formed. It isn’t moving.
The RTO Push Is Loud, Not Wide
The visible return-to-office campaigns are concentrated, not widespread. They cluster in legacy financial services, a few tech firms, and public-sector where lease commitments still drive agenda. Outside those pockets, the picture is different: hybrid is the median, not the exception, and full-remote roles attract more applicants per posting than equivalent on-site roles, even at compressed wages.
The reason mandated RTO feels louder is selection bias. Firms that issue mandates put out press releases. Firms that quietly run hybrid don’t. The signal-to-noise ratio favours the side that’s losing the argument.
Output Has Replaced Hours as the Unit of Account
Despite this, a deeper shift sits underneath the location debate. For a long time, the basic unit of professional value was time. You sold hours. You were managed on attendance. Your seniority compounded on tenure. Remote work was incompatible with that model — not because people couldn’t deliver, but because the management apparatus was built to measure presence.
That apparatus is being dismantled. Output-based metrics — closed tickets, shipped features, completed evaluations, sales numbers, and deal margin — are operational reality across the companies that actually pay for performance. Where the unit of value is output, location becomes a logistics question, not a strategy one.
For many professions, the era of trading hours for money is finished. The transition is not complete and not evenly distributed, but the direction is fixed.
The Independent Operator Is Becoming the Default Pattern
When the unit of account is output, the contract structure follows.
Independent contributors — fractional executives, specialist consultants, technical leads, advisors on equity — are absorbing work that used to belong to an FTE. Skilled-professional freelance work has been growing in double digits each year since 2022, with the steepest growth in roles previously considered untouchable: strategy, security architecture, finance.
The professional class is decoupling from the company. People with rare judgement are operating across multiple clients, owning their own tooling, and using AI to compress the operational overhead that used to require a back office.
Remote isn’t the cause. It’s the enabler. The cause is leverage.
The best people are leaving — not the company, but the model. Multiple clients. Own tooling. AI handling the operational noise that used to need a whole support function.
Location is a symptom. Leverage is the diagnosis.
The Home Stack Has Caught Up to the Enterprise Stack
For five years the case against remote was infrastructure. That argument has expired.
Fixed-wireless 5G now delivers gigabit-class throughput in the urban and suburban patches where most knowledge workers live. Consumer-grade displays, microphones, and lighting routinely outperform what most enterprises specify. Domestic broadband resilience — dual SIM failover, mesh networking, UPS-protected routers — costs less than a single corporate ergonomic chair. The home office, properly set up, is no longer the constrained version of the corporate office. In many cases it is the better one – mine is luxury compared to the office, and custom to my work patterns.
Meanwhile the enterprise estate is ageing. Open-plan floors retrofitted from pre-pandemic densities. Booking systems for desks that no one wants on Mondays or Fridays. The asymmetry is now reversed.
Shadow Stacks Are Building Capability Faster Than IT Is Building Policy
There is a quieter story underneath the headline numbers. It is the one the IT function inside most large organisations is trying not to admit.
The most capable workers are running their own AI stacks. Personal Claude or OpenAI subscriptions. Private Notion or Obsidian knowledge bases. Personal n8n or Zapier workflows wired into the work they get paid to do. Cursor and Windsurf instances tuned with personal context. None of it appears on the corporate risk register because none of it is licensed by the corporate. It exists in the gap between policy and practice… and the CSOC and DLP protocols the last blocker to a wave of changed delivery methods.
This is shadow IT, but that description undersells the phenomenon. What’s building is a shadow capability layer. The work product the employer receives is increasingly produced through tools, models, and data flows the employer does not control, has not approved, and cannot reconstruct if the person leaves. An evolution of BYOD on steroids.
For SMEs the issue is acute and the response is faster: many are now formalising what staff were already doing — issuing licences, defining data boundaries, treating personal AI use as an operational dependency rather than a compliance violation. Enterprise is slower, which means enterprise is widening the gap between what its workforce actually does and what its estate can see.
Sovereignty Is the Next Battleground
The shadow stack changes the centre of gravity for three questions large organisations have not yet answered seriously.
- Data ownership. When a senior consultant drafts a proposal in their personal Claude account, prompts it with anonymised client context, refines it across twenty iterations, and pastes the result into a corporate document — who owns the prompt history? Who owns the model context that shaped the output? Where does that conversation live, what are the legal rights, and what does the employment contract actually cover? The answer in most companies today is: no one has looked.
- IP and IPR. Personal AI workflows encode tacit judgement — the operator’s curated prompts, examples, system context, and refinement loops. That judgement is what the employer is paying for. But the encoded version is portable. It walks out with the operator. Most employment contracts predate this and don’t cover it.
- Information sovereignty. When the work product passes through US-hosted frontier models, the data leaves the geographic region. For UK regulated companies, public-sector work, OFFICIAL material, and anything touching protective services, the implications are not academic. They are the difference between a defensible audit and an enforcement action. In my own world — mission-critical, scrutinised — that gap between policy and practice is not a thought experiment. It is the daily operational risk no one wants to write down.
Where Are We Now?
Remote work has quietly won one argument, but thats largely irrelevant now. The more interesting question is no longer whether the promise of muffins, baked goods and a Friday beer bought people in, it’s what capability and shadow IT hat snuck in the door, and how people are working with it.
The companies that win the next decade will be the ones that recognise their best people are already operating like firms-of-one, and give them the licensed tools, the data agreements, and the sovereignty boundaries to do it cleanly inside the employment relationship.
The companies that lose will be the ones still arguing about badge swipes. Who miss the systemic changes in other user habits.
For more on artificial intelligence, autonomy, leverage, and the irreversible changes now reshaping work and modern life, explore the other posts or subscribe.
