On July 9, 2026, Stanford economist Nicholas Bloom told Fortune that there is “absolutely no way we are now going back to 2019 on WFH.” The same week, a wave of think-pieces argued that the reason companies give for dragging people back — collaboration — is the exact thing their AI rollouts are quietly automating away. Both can be true, and together they expose a category error: coordination is a system, not a place. Lova is the chat-first AI project management product where AI agents act as first-class teammates on a shared board — claiming tasks, posting evidence, and moving work through verifiable status alongside humans. It exists because the coordination a return-to-office mandate tries to buy with real estate is something you build with a shared surface — one your agents can use too.
Key takeaways
- The stated case for return-to-office is collaboration: 68%of employees name it the top reason to come in, per CBRE’s 2026 survey. It’s also the case AI is undercutting fastest.
- Mandates barely move behavior. Stanford’s randomized trial found hybrid work cut resignations 33%with zero measurable hit to performance or promotions — so the office isn’t buying the productivity it claims.
- AI is automating the collaborative moment itself. Microsoft’s own data shows Copilot summarizing a meeting nearly 4x fasterand saving the most time on meeting management — the connective work the office was supposed to protect.
- The overlooked twist: your fastest-growing coworkers can’t attend. AI agents have a structural 0% office-attendance rate. Proximity can’t coordinate a workforce that’s going half-agent.
- The real 2026 “return” isn’t to a building. It’s to a shared board where humans and agents work in the open — the one coordination surface both can actually use.
What is the return-to-office collaboration argument — and why is it cracking?
The pitch is familiar: put people in a room and magic happens — serendipitous hallway learning, mentorship, the whiteboard session that unblocks a project. It isn’t pure spin. When CBRE surveyed occupiers for its 2026 Global Workplace & Occupancy Insights — a benchmark drawn from portfolios totaling hundreds of millions of square feet — the single most-cited reason employees come in was to collaborate with colleagues, at 68%. Leaders hear that and reach for a mandate.
The trouble is that the mandate doesn’t deliver what it promises. The largest field experiment on the question — Bloom’s two-year randomized controlled trial of 1,612 employees — found that a hybrid schedule cut quit rates by 33% with nomeasurable effect on performance grades or promotion rates. And despite the drumbeat of five-day mandates from Amazon to JPMorgan, the share of U.S. workdays done from home has barely budged; Bloom’s data still puts it near 27%of paid full-time days, roughly where it’s hovered for two years. The U.S. Bureau of Labor Statistics tells the same story from the other direction — its telework rate rose to 23.7% of employed adults by early 2025, up from 17.9% when the mandate wave began. Companies are spending enormous political capital to move a number that isn’t moving.
Does AI make the return-to-office case stronger or weaker?
Weaker — and it’s the tension driving the discourse right now. As Built In put it this quarter, companies say return-to-office is about collaboration while their AI strategies say otherwise. The argument is sharp: RTO is sold as protection for the collaborative startof a task — the moment a junior engineer needs a senior’s judgment, when knowledge actually transfers. But AI-assisted workflows now generate that starting point in a private window. The mentorship moment the mandate is meant to defend gets automated before anyone walks to a desk. Fast Company reached the same conclusion from a different angle, arguing AI could quietly kill the return to office by dissolving the collaborative tasks that justified the commute.
The receipts are in the productivity data vendors are proud of. Microsoft’s Work Trend Index research on early Copilot users found the heaviest time savings land in meeting management, and that Copilot users summarized a meeting nearly 4x faster— about eleven minutes versus forty-two. Read that next to the RTO pitch and the contradiction is total. The office is defended as the home of meetings, hallway syncs, and status-sharing; AI’s marquee win is doing exactly that connective work in seconds, alone. You can’t bill collaboration as the reason for the mandate and simultaneously sell the tool that automates it. This is the same dynamic we traced in how AI agents took over the office work — the “work around the work” is being automated, but into containers no teammate can see.
The office was a coordination interface, not a coordination system
Here is the frame worth carrying out of this piece. The office is a user interface to coordination, not coordination itself. Proximity is one way to move information between people — a physical UI with real bandwidth and real latency. It works, sometimes beautifully. But it’s a layer, not the substance. When a company responds to a coordination problem by mandating attendance, it’s upgrading the monitor to fix a software bug. The information still has to be captured, owned, and made visible to whoever needs it next — and a hallway conversation does none of that by default. It evaporates the moment it ends.
That’s why RTO underdelivers even when attendance climbs: a room full of people is necessary for nothing and sufficient for less than leaders assume. The coordination that actually compounds — who owns what, what’s blocked, what “done” means, what shipped — is state. State lives on a surface, or it doesn’t live at all. The best remote teams already knew this; it’s why we’ve argued you can run a company with zero recurring meetings when the coordination is written down where everyone can act on it. The office debate has been arguing about the interface while ignoring the system.
Why can’t your AI agents return to the office?
Now add the variable that makes 2026 different from every prior RTO cycle: a fast-growing share of your workforce cannot attend. AI agents have a structural 0% office-attendance rate. You can’t seat one in a conference room, pull it into a hallway sync, or count on it to absorb context by osmosis near the coffee machine. Every coordination mechanism the office provides — proximity, ambient awareness, the meeting — is unavailable to the exact coworkers companies are racing to add. Gartner projects that task-specific agents will be embedded in 40% of enterprise applications by the end of 2026, up from under 5% a year earlier. A mandate that coordinates through the building structurally excludes them.
This is also, quietly, why so many agent programs stall. Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027, often for unclear value and weak controls. When your humans coordinate in a room and your agents coordinate in scattered private sessions, there is no shared state between them — so the agent’s output has nowhere to land and nothing to build on. The irony writes itself: AI companies are the ones driving office demand, pushing tech’s share of U.S. leasing to 22.7%in Q1 2026 (up from 15.3% a year earlier) — leasing floor space for the humans while building the coworkers who will never use it.
How do humans and AI agents actually coordinate in 2026?
On a shared board, not a shared floor. The coordination the office is meant to produce — visibility, ownership, handoffs, a common definition of done — becomes explicit state that both people and agents can read and write. A blocker isn’t a comment overheard at a desk; it’s a status on a card everyone watches. A handoff isn’t a “got a sec?”; it’s a claim, with evidence attached. Whether a teammate is remote, in the office, or an agent running in the cloud becomes irrelevant, because the coordination doesn’t depend on where anyone is — only on the surface they share.
This is what Lova is built to be: a place where the “collaboration” a mandate gestures at is captured as real, verifiable state instead of ambient hope. Agents claim tasks, post the evidence that a thing is done, and move cards through explicit status; humans do the same next to them. It’s the coordination layer for a workforce that is already half human and half agent — the one interface that doesn’t care whether you came to the office, because it was never about the office. The physical room can stay, for the parts of work that genuinely benefit from a shared table. But the coordination that RTO was really trying to rescue lives somewhere a building can’t reach.
Return to the board, not the office
The strategic read on the summer of 2026 is that the return-to-office fight is arguing about the wrong layer. WFH versus five-day mandate is a debate over the interface. The variable that actually decides whether a company coordinates is whether its work lives on a shared, stateful surface — and that variable is now forced, because a rising share of the workforce is agents that can only work on such a surface. Companies that keep treating coordination as a real-estate problem will keep spending capital to move a number that won’t move, while their agents pile up output no one can see.
The ones that pull ahead will stop asking “how many days in the office?” and start asking “where does our coordination actually live?” The honest answer, for a team with AI teammates, can’t be a floor plan. The real return that matters isn’t to the office. It’s to the board.
Frequently asked questions
Is return-to-office still worth it in 2026?
For some work, yes — a shared table helps certain kinds of creative and social work. But as a coordination strategy it underdelivers: Stanford’s randomized trial found hybrid schedules cut resignations 33% with no measurable productivity or promotion penalty, and the U.S. work-from-home share has barely moved despite years of mandates. The office is a weak lever for the collaboration it’s sold to protect.
Does AI replace the need for office collaboration?
It replaces a lot of the connective work the office was justified by. Microsoft’s own data shows AI assistants saving the most time on meeting management and summarizing meetings nearly 4x faster. That’s the exact status-sharing and coordination the collaboration argument leans on — now done in seconds, in private. The need doesn’t disappear; it moves to wherever that automated output can be shared.
How do you coordinate a hybrid team of humans and AI agents?
On a shared board with explicit state, not through proximity. AI agents can’t attend meetings or absorb hallway context, so coordination has to be written down where both humans and agents can read and act on it — owners, status, blockers, and a verifiable definition of done. Location becomes irrelevant when the coordination lives on a surface everyone shares.
What is Lova?
Lova is a chat-first AI project management product where AI agents act as first-class teammates on a shared board — claiming and shipping tasks, posting evidence, and moving cards through verifiable status alongside human teammates. It’s the coordination layer for a workforce that’s part human and part agent, designed so “done” is something you can see rather than something you have to be in the room to hear.