The occupancy gap is closing, and estimation is what got left behind

What the research found

JLL's Global Occupancy Planning Benchmark Report 2026 covers 84 organisations and 716 million square feet of portfolio across North America, Latin America, Europe, the Middle East, Africa and Asia Pacific. Global office utilisation sits at 56 percent against a target of 74 percent. That is an 18 point gap, and it is an improvement: a year earlier the gap was 25 points.

The reason for the improvement is the part worth reading twice. Ninety percent of organisations now apply utilisation data directly to space planning decisions, up from 70 percent in a single year. Improving the accuracy of that data has risen to the second highest priority for corporate real estate teams, behind only portfolio optimisation.

The work patterns underneath have also settled. Sixty two percent of organisations now require fixed in-office days, up from 49 percent, and 55 percent of employees attend three to four days a week, a rise of 19 percentage points year on year. Fully remote working has halved, from 18 percent to 10 percent. Where mandates exist, 64 percent of organisations report utilisation increases.

And where demand is going is visible in the space types being added: phone booths in 41 percent of portfolios, focus rooms in 30 percent, small meeting rooms in 29 percent.

What we see happening

Three years ago the workplace conversation was about whether people would come back. That question has been answered, and a harder one has replaced it: can you prove what your building is doing.

The shift from 70 to 90 percent in a single year is not a technology story. It is a credibility story. Utilisation numbers now travel further into an organisation than they used to. They land in board papers, lease negotiations, ESG reports and finance reviews, and they get read by people whose job is to find the weak assumption. A number sourced from bookings or badge swipes does not survive that room any more.

We also read the AI findings as a warning about sequencing. Over 70 percent of the organisations in the study have not piloted, optimised or scaled AI in workplace operations, and only 7 percent rate their own data capabilities as excellent. Forecasting demand, predicting peak days, automating building response: none of that works on a foundation of estimates. The organisations that will use AI for capacity planning in 2028 are the ones building a clean, measured occupancy baseline in 2026. There is no shortcut, because a model trained on booking data will confidently predict intentions rather than behaviour.

The space type data points the same way. Portfolios are being rebalanced toward small rooms, and you cannot make that call from a reservation system. Booking data tells you a room was taken. It does not tell you that an eight person room hosted a two person call, which is precisely the observation that justifies converting it.

What this means for how you run a building

The practical move is to separate two things that most organisations still report as one: what was reserved, and what was used.

iotspot Insights adds the measured layer beneath the booking layer you already run. Desk and room sensors record actual use per desk, zone, floor and hour, on the floor plans already configured in iotspot. That gives you a sharing ratio set on evidence, a cost per square metre that reflects real behaviour, and layout changes you can defend line by line.

From there the same signal does more than reporting. It drives climate through iotspot Smart Climate so conditioning follows occupancy rather than a timetable, and it feeds iotspot ESG Reporting so sustainability figures carry their measurement with them.

The honest caveat

JLL's benchmark is strong because it discloses what it is: 84 organisations, 716 million square feet, measured occupancy rather than self-report. It is also a benchmark of large, mostly multinational portfolios, so a single site in one country should treat it as direction rather than target. The most useful number in your own organisation is still the one you measure yourself, which is exactly the argument the report is making.