Two-thirds of full-time office workers in the Gensler Research Institute’s 2026 Global Workplace Survey say they are already fixing the office themselves. Gensler’s word is hacking: people adjusting their own workspaces to compensate for performance gaps the building never closed.
The improvisation is easy to picture, and it rarely looks like a capital project. A monitor dragged out of the glare. A quiet corner claimed before anyone else arrives. A jacket over a vent that will not quit. Those are the kinds of patches facilities teams walk past. They are not a coded breakdown of what the two-thirds specifically changed.
Keep the findings apart. Two-thirds is the broad hacking claim. Gensler did not find that two-thirds reworked lighting, temperature, or noise. Separately, Gensler found that one in four resorted to do-it-yourself fixes for ergonomics, temperature, or visual privacy. Noise and meeting-space availability, in Gensler’s wording, remain unresolved challenges.
The survey covered 16, 459 full-time office workers in 16 countries, collected from July 22 to September 12, 2025, at employers of at least 100 people who use an office at least some of the time. It is not a small-business sample, and it says nothing about people who never come in. It also does not report a productivity percentage.
A hot room and a full calendar are different problems
People hack the office because layout, lighting, temperature, and noise are still often set by taste, and by whatever the last headquarters tour praised. The building cannot tell demand from occupancy, or a hot room from a booked one. Until it can, the patches continue.
Three inputs answer three questions. Mix them into one score and the decision goes fuzzy. On Monday, pick a single gap: noise, meeting-room fit, heat, or visual privacy. Baseline that. Not all four.
- Sensors can show occupied room-hours against available room-hours, and room conditions while the room is in use. Before you act, check coverage, timestamps, and missing readings. A gap in the file is not a quiet Tuesday.
- Bookings show demand: booked versus available room-hours, and reservations with no check-in. A booking is a claim on a room. Nobody has to be in the chair. A full calendar and an uncomfortable room are different problems.
- Surveys capture reported comfort, distraction, and the kinds of space people say they need. Ask the same questions before and after. Compare response rates and attendance before you declare the change a success.
Space utilization, office attendance, and employee satisfaction belong in the same conversation. Collaboration analytics, the software that describes how people message and meet, answers a different question. It will not tell you whether the west-facing room is too hot at 3 p.m. An acceptable reading on a building system can sit next to a sweater on a chair. Readings and reported comfort are both real. You cannot swap one for the other.
The scenes teams already suspect (small rooms booked solid while large rooms sit empty, sun-facing rooms heating up after lunch, conference rooms held for one person on a video call) are hypotheses. Test them against the right evidence. Do not treat them as a shopping list.
The mix held. One template still will not.
Over the prior three years, Gensler found the mix of work relatively consistent. The average respondent spends more than half their time in the office, and 26 percent of the workweek somewhere that is neither office nor home: coworking, a client or vendor site, a third place, or business travel. Time across focus work, in-person collaboration, virtual collaboration, social connection, and learning held steady too.
Stable is not the same as uniform. How and where people work still differs by country, industry, age, and role. One office template was always the wrong brief. Copying it now only makes the mistake more expensive.
Gensler’s conclusion is that if learning and social connection keep mattering as AI changes work, the office becomes an investment in well-being, learning, and collaboration. That is Gensler’s conclusion, not a measured return. It is a clearer brief than adding desks and hoping people collide productively.
What the survey does connect to a workplace people rate as effective for learning is more concrete. In regressions that accounted for role, industry, and country, among other factors, Gensler linked that judgment to conditions a facilities lead can actually inspect. Three of them: manageable noise, easy access to a place to focus, and easy access to somewhere to rest and recharge. Asked about the future office, workers also put wellness and access to nature high on the list, including outdoor space and natural light, along with rooms that feel productive, creative, and inspiring. Rankings shift by country, industry, AI use, and learning orientation. There is no global ranking to photocopy.
Match the room to the work
Open plan can help a casual conversation and wreck the hour after it. A noise reading only means something if you know which room, which activity, and which hour it came from. A measurement in an empty office does not explain complaints during a packed meeting block. Acoustic panels, a designated quiet area, and a clearer split between collaborative and individual space are worth testing in a few comparable rooms over the same period. Roll them out on a hunch and you will not know what helped.
Background sound may suit a reception or a lounge. Concentration space usually wants quiet. That is a judgment about fit. Individual focus, virtual meetings, project work, and casual conversation do not want the same room, and departments do not want the same week. The aim is a range people can choose, not one atmosphere forced to please every task.
Frequent booking of meeting rooms for solo calls is a signal to investigate, not proof that you need a row of phone booths. Poor acoustics or unreliable equipment somewhere else can produce the same reservation pattern. When needs conflict, the data will not name a winner. You can automate the readings. You should not automate the verdict. Someone still has to decide, and the people who use the space should be in that review before the furniture arrives.
Let the experiments proliferate, and let the people who are working in the spaces feel ownership over them.
Ethan Bernstein put it that way in remarks to the Harvard Gazette on open-office research. The remark is not a measured outcome. Treat it as the working rule. Record conditions first. Change what the evidence supports. Measure again the same way. Then read the result with the rest of the month in view: attendance, a shift in working arrangements, a quiet stretch in the business, no-shows, sensor gaps, and how many people bothered to answer.
If comfort is the goal, ask whether the room feels better. A quieter complaint log is a clue. It is not, by itself, proof the redesign caused it.
A space study is not a person study
If people think the space study is a person study, the feedback dries up and the hacking goes private. More data does not have to mean closer watching. The useful question is how spaces function, not where a named person walked.
Three rules. Use the least detailed data that answers the facilities question. Tell people what will be collected, who can see it, and when it will be deleted. Do not cut results so finely that a team of six can recognize itself. Grouping by space is not the same as anonymity. Bookings are the easy place to over-collect. If the analysis needs named reservation records, say so, and ask whether it actually does.
Questions worth asking before you move a wall
- What does Gensler’s “two-thirds” figure actually measure?
It is the share who report hacking their workspace to compensate for performance gaps, not a finding that two-thirds specifically changed lighting, temperature, or noise. The one-in-four figure, on do-it-yourself fixes for ergonomics, temperature, or visual privacy, is a separate result.
- Does better workplace data mean monitoring individuals?
It should not. Use the least detailed occupancy, booking, and survey data that answers a space question, say what is collected and when it is deleted, and remember that grouping results does not automatically make people unidentifiable.
- How do you know a change worked?
Repeat the same measures after a limited pilot, compare similar spaces over the same period, and check attendance, response rates, and no-shows before you credit the redesign. A quieter complaint log is a clue, not a cause.
- If meeting rooms are always booked for solo calls, do you need more focus pods?
Not automatically. That pattern is a reason to look. Bad acoustics or unreliable equipment elsewhere can push the same behavior into conference rooms.
- Can you claim a productivity gain from this?
Not from this evidence. Gensler’s survey does not report a measured gain in productivity, retention, cost, or satisfaction, and a baseline-and-remeasure habit does not supply one on its own. If comfort is the goal, ask whether the room feels better.
- Which single gap will you baseline before you move a wall?
Pick one: noise, meeting-room fit, heat, or visual privacy. Record it with the evidence that actually speaks to it, change only what that baseline supports, and leave the other three for the next round.
The office is still where more than half the week happens, on average. Noise and meeting space are still unresolved. A quarter of workers are improvising ergonomics, temperature, and visual privacy on their own. AI can change how that week is spent. It will not, by itself, cool the west-facing room or free a meeting space that was never the right size.
The building will not catch up to the work by accident.