Most commercial real estate decisions about office space are built on occupancy data that is, at best, incomplete. Badge swipe systems tell building operators that someone entered a floor. They do not capture dwell time, conference room utilization, which collaborative zones are actually generating activity, or whether thermal comfort conditions in a given quadrant are driving people out by mid-afternoon. The result is a persistent gap between what operators believe about their buildings and what is actually happening inside them.
Honghao Deng, co-founder of BUTLR, has spent years building technology to close that gap. The company, a spin-out from MIT’s City Science Group, deploys thermal infrared sensors that generate spatial intelligence data across office campuses, university buildings, healthcare facilities, and data centers. In a recent episode of The Real Finds Podcast, Deng outlined what that data is actually showing about how physical space is used, what landlords and occupiers can do with it, and why the commercial real estate industry is still underestimating what its buildings are telling it.
The Limits of Badge Data
Badge swipe systems were designed for security, not analytics. They log access events at building entry points or on individual floors, but they cannot answer the questions that matter most to a facilities manager or a leasing strategist. Did that person stay for four hours or four minutes? Did they use a conference room or sit at a desk? Did they collaborate with a team from another department, or take a solo call at their workstation? When an entire floor swipes in on the same morning because there is free lunch in the break room, the system reads full occupancy.
As we covered in our analysis of why office space efficiency is broken, the gap between reported occupancy and actual utilization is consistently large across commercial portfolios. Research suggests that spaces are underutilized by 50 to 70 percent even in organizations that consider themselves fully back in the office. Badge data not only fails to capture that gap; it actively obscures it by creating the impression of full engagement from a fundamentally coarse input.
Dwell time, activity type, zone-specific utilization, and microclimate conditions require a different kind of sensing infrastructure. The reason that infrastructure has been slow to arrive, Deng argues, is not primarily technological. It is structural. Buildings have not historically been treated as products that generate feedback. They have been treated as containers.
Thermal Sensing as a Privacy-First Baseline
BUTLR’s sensors operate on infrared thermal data rather than visible-light imaging. They produce what Deng describes as thermal blobs: temperature readings from the environment that can be interpreted, through specialized machine learning models, into location data, headcount, dwell time, posture, and activity state, without capturing anything resembling a photograph or video feed.
This matters for reasons beyond the obvious legal ones. Organizations evaluating sensor-based occupancy monitoring frequently stall on privacy concerns, particularly in workplaces where employee trust is already a sensitive variable. A thermal sensor produces no footage. It cannot identify individuals. It cannot be repurposed for surveillance. That distinction has allowed BUTLR to deploy across environments, including restrooms, private offices, and patient care settings, where camera-based systems would be neither acceptable nor practical.
The underlying sensor modality is not new. Passive infrared motion detectors have been standard building components for decades, the same kind that trigger occupancy-activated lighting. What BUTLR has done is increase the spatial resolution and layer a specialized AI model on top of that familiar sensor category to extract far more actionable intelligence from data that has always been present in the built environment.
The third-generation sensor Deng described takes that further, using an opaque cover that blocks all visible light entirely, making it architecturally indistinguishable from a standard ceiling fixture while remaining fully functional. For organizations that have resisted occupancy sensing on privacy grounds, this is a meaningful threshold.
What the Data Shows About Conference Rooms
One of the clearest findings from BUTLR’s platform data is that large conference rooms are dramatically underutilized, running at occupancy rates of 20 to 30 percent across many enterprise clients. That figure consistently surprises organizations that believe, based on booking system data, that their conference rooms are perpetually overbooked.
The discrepancy is structural. Calendar booking systems show reservations, not presence. A room blocked for twelve people and actually used by two still registers as scheduled and occupied. Without dwell time data linked to actual headcount in the room, facilities teams have no mechanism to distinguish between those outcomes. Real estate strategy ends up being built around booking cadence rather than genuine utilization, leading organizations to add meeting room inventory they do not need while leaving unbooked collaborative zones empty.
BUTLR’s platform has gone further than identifying the gap. Using occupancy data as input, the company demonstrated a generative AI workflow in which a model analyzed actual utilization patterns and produced floor plan redesign recommendations without human prompting. The output converted underused large conference rooms into smaller huddle spaces and open collaborative zones, with suggestions that reportedly matched both functional requirements and aesthetic coherence. The intelligence was in the data; the AI translated it into actionable spatial recommendations.
With 1.5 billion frames of occupancy data generated across the platform each day, the statistical foundation for those recommendations is not anecdotal. It reflects aggregate patterns across a wide range of building types, geographies, and organizational configurations.
Energy, Maintenance, and the Operational Case
The business case for spatial intelligence does not begin with AI-generated floor plans. It begins with energy and maintenance, where the return on investment is immediate and measurable.
Buildings account for approximately 40 percent of global energy consumption, according to the International Energy Agency, a figure that puts even modest efficiency gains at significant scale once applied across a portfolio. The default operating logic of most building HVAC systems is time-based rather than occupancy-based: conditioning runs on a schedule regardless of whether anyone is present. The result is a system cooling an empty floor all afternoon and then struggling to recalibrate when occupancy spikes unexpectedly.
Tying HVAC actuation directly to real-time occupancy data is among the most straightforward applications of spatial intelligence, and one of the most impactful per dollar of sensor investment. When a zone is empty, the system does not condition it. When occupancy increases, the system responds. Energy savings compound at the portfolio level, and the comfort conditions that influence commercial property management outcomes, tenant satisfaction, lease renewal decisions, and maintenance call volume, improve alongside them.
Maintenance scheduling produces a parallel dynamic. A case study Deng cited with a major facility management firm showed that scheduling restroom cleaning based on actual traffic data rather than fixed time intervals generates approximately $1,400 in annual labor savings per restroom. A university campus operating 500 restrooms converts that per-unit figure into roughly $700,000 in annual operational savings from a single use case, using a single sensor per doorway. That math scales directly to any large commercial building, campus development, or managed portfolio.
The Landlord Intelligence Layer
The implications of spatial occupancy data extend beyond the occupier side of the relationship. For building owners and investors, real-time utilization data creates the ability to model tenant health in ways that go far beyond rent payment history and calendar booking data.
A tenant whose occupied zones show steadily declining activity over a twelve-month period is a retention risk that badge data would not surface until lease renewal discussions were already underway. A tenant whose cafeteria and collaborative zones are running at high utilization while private office space sits half-empty may be a candidate for a proactive rightsizing conversation, one the landlord initiates with evidence rather than waiting for the tenant to raise it. In both cases, the landlord holding utilization data has strategic optionality that the landlord without it does not.
This is the direction of travel for commercial real estate portfolio management: from reactive to predictive. As we outlined in our post on preparing commercial real estate portfolios for AI disruption, the operational floor for what tenants expect from a well-managed building is rising. Owners who are building data infrastructure now will compound an information advantage that will become more consequential as the market continues to bifurcate between buildings that know how they are being used and buildings that do not.
The Autonomous Buildings Thesis
Deng frames the long trajectory of spatial intelligence through an analogy to autonomous vehicles. Early driver-assist systems did little more than maintain highway speed and following distance. The capability of those systems today, navigating city streets, handling intersections, responding dynamically to pedestrian and vehicle behavior, is a direct function of the data layer that accumulated over years of real-world deployment. Buildings, in his framing, are on the same development curve, just earlier in the timeline.
A building with a dense, reliable spatial data layer can begin to self-regulate: adjusting environmental conditions, anticipating maintenance needs, generating space reconfiguration recommendations, and flagging utilization anomalies without requiring a facilities team member to manually interpret a report. That is not a speculative future state. The functional components are operating in BUTLR deployments today. What the industry is still working through is the organizational shift required to treat buildings as intelligent, responsive products rather than fixed containers with mechanical systems running on timers.
The data center sector offers a useful leading indicator. Data centers are among the fastest-appreciating assets in commercial real estate precisely because they are not buildings in the traditional sense. They are operational platforms where asset value depends entirely on performance metrics: uptime, thermal management, power density, and adaptability to successive hardware generations. The spatial intelligence layer being built into office campuses moves general commercial real estate toward a similar value model, one in which what a building knows about itself begins to factor into what it is worth and how it competes for tenants.
For a closer look at what the physical office offers that no data layer can replicate, our analysis of real value in the in-person workspace covers the research on collaboration, creativity, and why the market is separating sharply between buildings that justify the commute and those that do not.
Van Vlissingen and Co. advises office tenants, building owners, and commercial real estate investors across Greater Chicago and the broader Midwest on space strategy, leasing, and portfolio decisions. As the market moves toward data-informed space management, our team helps occupiers evaluate how utilization intelligence should factor into their next lease and helps landlords assess how to position their assets competitively. Contact our team of Chicagoland commercial real estate agents to discuss how this shift applies to your portfolio or your next real estate decision.