A decade from now, the building you walk into will look almost exactly like the one you walked into today. The glass, the lobby, the lift, the carpet. The architecture won't have changed. The intelligence underneath it will have.
In the same decade, two demand pressures will land on the world's commercial real estate at the same time. Two billion more city dwellers will need spaces to live, work, study, queue, eat and pass through. And the same buildings - according to the International Energy Agency - have to cut 97% of their direct CO₂ emissions by 2050 while doing it. The two pressures pull in opposite directions: more people, less energy, the same floors. The only way through that compression is optimisation. The only way to optimise is to measure.
This post is about the direction building optimisation is going - and why the next ten years are going to belong to the operators who treat occupancy data as the currency, not the byproduct.
01 · THE TWO PRESSURES, IN NUMBERSThe two pressures, in numbers
It helps to look at the numbers up front, because the inflection isn't subtle.
Urban demand. United Nations forecasts say 67.2% of the world's population will live in cities by 2050. The urban population grows from 4.61 billion in 2023 to 6.68 billion by 2050 - 982 million new city dwellers, more than half of whom land in seven countries (India, Nigeria, Pakistan, DRC, Egypt, Bangladesh, Ethiopia). Material consumption to build the floors they need rises from 41.1 billion tonnes a year in 2010 to roughly 89 billion tonnes by 2050.
Energy. Buildings account for around 30% of global final energy consumption. HVAC alone consumes up to 44% of on-site energy in a commercial building. The IEA's net-zero roadmap requires 85% of the global building stock to be retrofitted to zero-carbon-ready standards by 2050 - against less than 1% today. Heat-pump installations have to scale from 1.5 million units per month today to 5 million by 2030 and 10 million by 2050.
The data layer. Forecasts of IoT density in commercial buildings project 4.12 billion connected devices by 2030. The smart-cities market into which those buildings plug is projected to grow from USD $952 billion in 2025 to USD $6.31 trillion by 2034 (Fortune Business Insights, 23.2% CAGR). The intelligent-building-automation subset alone is projected at USD $191.1 billion by 2030.
Hybrid work. Office utilisation in 2025 reached 53%, up from 38% in 2024 and 35% in 2023, with peak utilisation averaging 80%, per GlobeSt. Two-thirds of corporate real estate leaders expect portfolio contraction in the next three years, with hybrid work cited as the primary driver. And critically: 90% of organisations track utilisation, only 7% rate their data quality as excellent.
That last data point is the real signal. The measurement layer is broken even where it exists. That is the gap the next decade of smart-building optimisation has to close.
02 · THE OLD PLAYBOOK: SCHEDULES DESIGNED FOR THE WORST DAY OF THE YEARThe old playbook: schedules designed for the worst day of the year
For most of the last fifty years, building optimisation has been a scheduling exercise. The mechanical systems run to a calendar. The HVAC schedule is set for the busiest day. The lighting comes on at 7am because that's when the cleaners arrive. The lifts run dispatch profiles tuned to the busiest week of the busiest quarter of the busiest building in the portfolio. The schedule is conservative on purpose - because the cost of being wrong (a meeting room at 28°C, a stairwell in the dark) is far more visible than the cost of being right (the energy bill at the end of the quarter).
The result is exactly what you would expect. Commercial buildings worldwide run their plant against a worst-case profile they hit perhaps thirty days a year. The other 335 days, the plant runs hot for an empty floor. That is the baseline waste the next decade has to remove.
The old playbook fails in three places at once:
- It misses the variance. A floor that runs at 78% on Tuesday and 31% on Friday is operated identically. The schedule does not flex.
- It misses the local detail. Aggregate badge-swipe data tells you how many people entered the building. It does not tell you which floor they ended up on, or which side of the floor.
- It misses the timing. A building's busiest two hours of the day - the post-lunch return, the late-afternoon meeting block - rarely fall when the plant is configured to manage them.
Each of those misses is a measurement problem. None of them is an HVAC problem. The fix is not better mechanical equipment. The fix is a real-time signal of how the building is actually being used.
03 · THE NEW PLAYBOOK: OCCUPANCY AS CURRENCY, NOT BYPRODUCTThe new playbook: occupancy as currency, not byproduct
The shift the next decade requires is a reframing of what occupancy data is for. In the old model, occupancy data was reporting - quarterly utilisation studies, badge-swipe summaries for HR, ad-hoc walk-throughs by facilities. The data lagged behavior by weeks, and it lived in spreadsheets that nobody outside facilities ever opened.
In the new model, occupancy data is a control signal. It drives the HVAC schedule, the lighting state, the cleaning rota, the lease decision, the meeting-room release, the egress route. It drives the energy bill and the safety report and the sustainability submission. It is the input to every operational decision the building makes, in real time, every day. The dashboard isn't the deliverable. The dashboard is the byproduct of a measurement layer that has finally caught up to the systems it informs.
Two architectural choices make that shift possible.
Edge inference
The first choice is moving computer-vision inference onto the sensor. Until recently, a building that wanted vision-based occupancy data uploaded video to the cloud and ran the model there. That worked, in the way that a Rube Goldberg machine works. It produced numbers, but it also produced a privacy problem, a network problem, and a compliance problem that grew faster than the deployment did.
Edge AI inverts that. The model runs on a chip inside the sensor. The raw video never leaves the device. Only the anonymous count - "Zone B currently holds 24 people, density 0.47, dwell median 2.3 minutes" - travels north. Every problem the cloud-upload model created disappears, because the data the platform was uploading was the problem.
By 2030, this architecture is the only one that will pass procurement at scale. Three trends drive that:
- Privacy regulation. Australia's 2025 statutory tort for serious invasions of privacy, the EU AI Act's high-risk classification for biometric systems, and equivalent legislation in California and Singapore have moved the burden of proof onto the operator. A camera that uploads video has to prove it is not retaining identity-bearing data. A camera that does not upload video cannot retain it, because it never had it. The difference is structural and an auditor can read it from the architecture diagram.
- Network economics. A 4K video stream off a single sensor is roughly 25 Mbps. A count payload from the same sensor is a few kilobits per minute. Multiplied across the 4 billion sensors projected to be deployed by 2030, the network savings of edge inference are an order-of-magnitude argument, not a marginal one.
- Cyber attack surface. A centralised video archive is, in the language of the SecurityBrief report on the 2026 AI arms race, a honey pot. A distributed, anonymised sensor network is not. The architecture that keeps the privacy answer load-bearing is the same architecture that reduces the breach footprint.
Live, per-zone density
The second choice is structural too. The building of 2030 does not have one occupancy number. It has dozens. Per zone, per minute, per day, calibrated against a capacity that the safety planner signed off on.
A per-zone, per-minute density signal is the input every other building system has been quietly waiting for. HVAC plant can be staged to follow the load rather than running flat. The reports layer can quantify what right-sizing the schedule against actual occupancy saved per quarter. The alerts layer can fire when a stairwell density crosses the egress threshold rather than relying on a fire-warden walk-through. The Talk-to-your-data assistant can answer "how was Library Floor 4 used last Friday" without a SQL query.
That layer is what turns the 30% of global energy that commercial buildings consume into a curve that can be flattened, not a number that can only be reported.
04 · WHAT THE DECADE ACTUALLY OPTIMISES FORWhat the decade actually optimises for
The headline goal is energy and emissions. The detail is more interesting.
Right-sized HVAC
The IEA's net-zero pathway ascribes 70% of building-emissions reduction to energy efficiency and electrification, with the remaining 30% coming from behaviour change and on-site renewables. Inside that 70%, right-sized HVAC is the single largest lever. A commercial building's HVAC profile against real occupancy returns 8% to 18% on the energy bill in the first year and a proportional emissions reduction. Heat-pump electrification (driving the IEA's 1.5M-to-10M heat-pump trajectory) compounds the saving by improving the conversion efficiency at the same time as the schedule cuts the demand.
This is the lever any individual building can pull. It does not require new construction, new mechanical equipment, or a new tenant. It requires a measurement layer that the plant can trust.
Floor right-sizing
The hybrid-work reset has converted office floors from utilisation-stable assets to utilisation-variable ones. The same financial-services firm referenced in OfficeSpace's space-optimisation analysis audited a portfolio over 90 days and identified a 33% footprint reduction at the next lease window. At up to $11,000 per employee in annual savings, the financial case is decisive. The blocker has been data quality, not architecture - and the platforms that solve the shadow-vacancy gap with anonymous occupancy data, not badge swipes, are the platforms that win.
Egress and density safety
As urban demand pushes density up, the cost of a static egress plan rises with it. Civic squares, transit interfaces, stadium concourses and library forecourts all face the same trade-off: more people, the same physical exits, a higher cost of getting routing wrong. Live density routing - the kind of layer a stadium uses to keep an egress envelope inside twelve minutes - becomes a civic obligation, not a venue luxury. The same architecture serves both.
The amenity flywheel
The new currency of office attendance is the amenity. The cafeteria queue, the lobby flow, the breakout zone density - they decide whether the workforce comes in. Pre-2020, that was a soft signal. Post-2020, it is a measurable one. The optimisation of corporate amenities - cafeteria queue alerts, lobby pacing during peak ingress, breakout-zone availability - has gone from nice-to-have to determinant of the hybrid attendance curve.
05 · THE TRANSIT INTERFACEThe transit interface
Smart buildings do not optimise in isolation. The next decade integrates the building with the transport interface in front of it. The post-match egress that releases 60,000 people onto a 200-person platform is the most extreme version of this problem, but it is not the only one. A library that empties at 6pm onto a tram stop. An office tower whose lift lobby fills the foyer two minutes before the bus arrives. A stadium that paces concourse release rates to match the next train, so the transport-hub seam becomes a routing decision instead of a crush risk.
By 2030, the building's release rate and the transit interface's absorption rate are exchanged across a public protocol. Both numbers are anonymous - per-minute densities and projected throughput, not identities. The seam is what turns a smart building into a participant in the smart city, not an island inside it.
06 · THE BLOCKER: 90% MEASURE, ONLY 7% TRUST THE DATAThe blocker: 90% measure, only 7% trust the data
The most consequential statistic in this whole picture is that 90% of organisations track utilisation and only 7% rate their data quality as excellent. Every other forecast in this post depends on closing that gap.
The reason the gap exists is that most utilisation measurement is built on the wrong primitives. Badge swipes capture entry, not occupancy. Manual surveys capture the day the surveyor walked the floor. Booking-system reservations capture the intent to use a room, not the use of it. Each of those primitives is a proxy for the thing the operator actually needs to know - how many people are in this zone right now - and each proxy degrades the signal further from the truth.
Privacy-first computer vision is the primitive that converts the proxy into a measurement. It counts what is actually there, in real time, anonymously. The 7% who currently trust their data are the ones whose measurement layer has caught up to the systems it informs. The 83% in between are the addressable market the next decade closes.
07 · WHAT THE OPTIMISED BUILDING OF 2030 LOOKS LIKEWhat the optimised building of 2030 looks like
If you sketch the building of 2030 as a stack, six layers shape it.
- The mechanical layer. Electrified HVAC, predominantly heat-pump driven, retrofitted to zero-carbon-ready standards across 85% of the building stock by 2050. Plant capacity sized for the building's actual peak, not the worst-case peak.
- The sensor layer. Edge-AI sensors at every zone, emitting anonymous counts per second. No video uploaded, no identifying data produced. Privacy compliance is structural.
- The data layer. Real-time per-zone occupancy and density, normalised against calibrated capacity, exposed as a control signal not just a report.
- The integration layer. BMS, Slack, Snowflake, REST - the existing operations stack consumes the new signal natively. The data layer fits the stack rather than replacing it.
- The decision layer. Right-sized HVAC schedules, dynamic egress routing, threshold-driven alerting, sustainability reports that walk into the assurance pack without rewriting the methodology.
- The civic layer. The building participates in the transport-hub seam, the council density model, the regional energy-demand signal. It is a node, not an island.
The same architecture serves all six. That is the design point worth taking from the next decade: the building of 2030 is not six stacks bolted together. It is one stack, one measurement layer, six surfaces it informs.
08 · THE DIRECTION OF TRAVELThe direction of travel
Building optimisation in the 2030s isn't about more cameras or smarter dashboards. It is about the slow, structural shift from buildings that report on themselves to buildings that operate from their own signals. The dashboards become byproducts. The schedules become consequences. The energy bills bend. The emissions curve bends with them.
Occivar is building toward the role of the intelligence layer in that stack - the layer that produces the count, anonymises it on the device, and surfaces it as the operational truth every other system reads from. The platform exists because the alternative architectures collapse under the privacy weight as soon as the regulator looks at them. Privacy-first edge AI is the only architecture that scales to a trillion-dollar smart-city build-out, because it is the only one whose privacy answer survives the next twenty regulatory cycles.
The buildings that win the next decade are the ones whose operators stop guessing. The platforms that win are the ones whose data the operators trust. The cities that win are the ones who treat the building's measurement layer and the transit's measurement layer as components of the same civic infrastructure.
If you operate a building, a portfolio, or a precinct and the gap between what you currently measure and what you need to measure is the conversation worth having, request a demo or read about how the privacy architecture was built to make the answer load-bearing in the first place.