Stand in the center of a 1970s office tower on a Tuesday afternoon and look toward the windows. They are sixty feet away. You cannot see them clearly because the fluorescent grid above you was designed to compensate for what the architect already knew when he drew the floor plate: daylight will never reach where you are standing. It did not need to. Office workers sat in cubicles under artificial light for eight hours, and nobody filed a code complaint about the experience because the building codes governing commercial occupancy do not require natural light in workspaces the way residential codes require it in bedrooms and living areas.
Now an algorithm has looked at this building, run its measurements through a model trained on hundreds of prior assessments, and decided it can become apartments.
Gensler's Conversions+ tool has assessed more than 1,300 office buildings across North America and found that roughly 25% are suitable for residential conversion. It evaluates 150-plus building factors, from floor plate depth and window spacing to elevator counts and HVAC configurations, and produces a feasibility score in hours rather than the weeks a traditional architectural assessment requires. Cities are already running it: Calgary fed its entire downtown inventory through the tool and identified enough conversion candidates to produce 24% more residential units than conventional planning had projected. Fast Company named Gensler one of 2024's most innovative companies partly because of this algorithm, a recognition that was not undeserved given how dramatically it compressed what used to require weeks of billable architectural assessment into something that finishes before the afternoon espresso arrives.
Nobody disputes that the tool is fast, or that the math behind it is sophisticated. What concerns me is what happens to the space the algorithm cannot fix.
The Light Line
Residential building codes in most U.S. jurisdictions require habitable rooms to have windows providing natural light. San Francisco requires a glazed area equal to at least one-eighth of the room's floor area. New York's Multiple Dwelling Law requires every room used for living or sleeping to have a window opening to the outdoors or to a compliant court or yard. These are not suggestions. They are law, and they exist because decades of public health research demonstrated that chronic deprivation of natural light degrades human health in measurable, dose-dependent ways.
In a typical post-1950 office building, the floor plate runs 80 to 120 feet deep from exterior wall to exterior wall. Place apartments along the perimeter, each requiring windows in every habitable room, and you consume roughly 25 to 30 feet of depth on each side. That leaves a core of 20 to 70 feet that cannot legally contain a bedroom, a living room, a kitchen with an eat-in area, or anything else the code classifies as habitable space.
San Francisco architect Charles Bloszies, who has studied this problem in detail, measured the residual dark space in a sample of San Francisco office buildings being evaluated for conversion. It ranged from 12 to 19 percent of total floor area, which means that for every five floors of converted apartments you are financing, somewhere between half a floor and a full floor will never produce a dollar of residential revenue. Not 12 to 19 percent of one floor, but of the entire building.
That is space you paid to acquire, space you will pay to retrofit, space that will carry property taxes and insurance premiums in perpetuity, and space where nobody can legally sleep.
What an Algorithm Sees and What a Resident Feels
Gensler's 150-factor model knows about floor plate depth. It knows about window-to-core distance and incorporates these measurements into its feasibility scoring, and buildings with excessively deep floor plates score lower or fail outright. The tool is not naive about the problem. But suitability and livability are different questions, and the gap between them is where the conversion boom will eventually produce its most uncomfortable outcomes.
A building that scores "suitable" might yield 200 apartments in which every habitable room meets code requirements for glazing, ventilation, and egress. Those apartments might still have hallways that run 40 feet from the elevator to the front door through corridors that will never see a photon of sunlight that did not first pass through a light tube, a borrowed-light transom, or an electrically powered panel designed to simulate a sky that exists five stories above the ceiling. The common areas, the mail rooms, the fitness centers carved from the building's dark core will function adequately under artificial illumination. The parking garage converted from an underground level will be a parking garage that looks exactly like every other parking garage. None of this violates code.
It violates something else. Architecture, when it works, creates spaces that make people feel held, spaces that participate in the hours and seasons of the life unfolding inside them. Light is not an amenity, not something you list between the dishwasher and the in-unit laundry. It is the medium through which a room communicates with the person inside it, telling them what time of day it is, what season, whether a cloud just passed or the sun broke through, whether the world outside is participating in their experience of being home. A bedroom that meets the minimum glazing ratio and nothing more, situated at the end of a 40-foot windowless corridor in a building whose bones were drawn for a different species of occupancy, is not a home in any way that matters beyond the legal definition. It is a compliant enclosure.
The Money Gap the Algorithm Cannot Close
Even when the architecture cooperates, the economics often do not. A Brookings Institution study partnered with Gensler, HR&A Advisors, and Eckholm Studios examined 18 real office buildings across six U.S. markets, produced hypothetical unit yields for each, and ran the financial models.
In five of six markets, no conversion was economically viable without public policy intervention. Five of six.
Conversion costs run $500 to $600 per square foot according to CommercialEdge, a number that already makes most municipal finance directors wince when they see it next to the achievable rents in their market. Gensler argues this is 30% lower than new construction, and in expensive coastal markets that comparison holds. But "cheaper than new" is not the same as "profitable," and the Brookings team demonstrated that the subsidy gap between conversion cost and achievable rents remains large enough in most markets to require tax abatements, zoning relief, or direct public funding before a developer can make the numbers work.
Factor in the dark space penalty, the floor area that produces zero residential revenue because no one can legally sleep there, and the numbers shift further still. If 12 to 19 percent of your acquired floor area produces zero residential revenue because it cannot legally contain apartments, your effective cost per rentable square foot rises by 14 to 24 percent above the headline conversion number. On a $550-per-square-foot base, that pushes effective cost to $627 to $681 per square foot of space someone will actually pay rent to occupy. The algorithm scored the building as feasible at $550. The architecture delivered it at $681.
Where Dark Space Becomes Something
Bloszies and his team have cataloged possible uses for the dark core. Storage units, which tenants will pay $150 to $300 per month to access and which produce revenue reliably but not at a rate that offsets the acquisition cost of the square footage they occupy. Shared amenity spaces like gyms and lounges, which add value to surrounding apartments but generate no direct revenue. Ground-floor retail, particularly grocery stores in food-desert neighborhoods where the conversion replaces a dead office building with both housing and daily necessities. Mechanical rooms for the building's new residential HVAC systems, which demand far more distributed infrastructure than the centralized commercial systems they replace.
Some proposals border on speculative territory, like autonomous vehicle staging areas, anticipating a future where residents' self-driving cars park themselves in windowless garages and retrieve themselves on command. Vertical farming installations, which require no natural light and consume precisely the kind of controlled, interior-accessible space that dark cores offer in abundance. Bloszies calls these "oddball" ideas himself, acknowledges that the honest proposals for unlightable space usually sound a little strange, and notes that the really polished ones tend to be the ones hiding bad economics underneath the rendering.
New York's approach is more pragmatic: the city's zoning allows office use within residential buildings, so a dark-core unit can legally serve as a home office, a second room that does not require a window because its designated use is not habitable occupancy. San Francisco prohibits this arrangement entirely, explicitly banning most live-work configurations in residential buildings under its current planning code. Whether that restriction survives the current conversion wave is an open question, but as of today it means SF's dark space problem is harder to solve than New York's by the width of a single zoning clause.
55 Broad Street and the Exception That Proves the Cost
The conversion everyone cites is 55 Broad Street in Manhattan's Financial District. Thirty stories. 571 luxury rental units. $220 million in financing from Banco Inbursa, arranged by JLL. LEED-certified. Rooftop pool. Self-contained HVAC. Metro Loft Management and Silverstein Properties as co-developers, with CetraRuddy as architect.
The per-unit cost: $385,000.
That number buys you a LEED-certified luxury apartment in lower Manhattan with proximity to transit, which sounds like a reasonable proposition until you realize it also buys you a building whose ground-floor retail program, amenity package, and common-area design were all shaped by the need to absorb dark-core space that the algorithm told the developer existed but could not tell the architect how to make beautiful. CetraRuddy solved it, being among the best adaptive-reuse firms in the country. The question is whether the 90,300 units in the 2026 pipeline will all have architects of that caliber, or whether most will have developers running Conversions+ on Tuesday and hiring the cheapest design-build firm on Wednesday.
Trust and Its Limits
A First American Data & Analytics study conducted with DealGround found that 66% of commercial real estate professionals now use AI tools weekly or daily. Only 5% trust those tools enough to base real deal decisions on their output alone. Fifty-three percent use AI exclusively for support, and 17% require heavy verification before acting on anything the model produces.
Those numbers describe an industry that has adopted the technology and does not yet believe it. For conversion feasibility specifically, the skepticism is well warranted. An algorithm can tell you that a building's floor plate depth, window spacing, and structural grid permit a residential layout. It cannot tell you whether the resulting apartments will feel like places worth inhabiting. It cannot tell you whether the dark core that remains after you carve out the apartments will produce enough ancillary revenue to offset its carrying costs. It cannot tell you whether the city council will approve the zoning variance you need to make the home-office workaround legal.
It tells you the building can be converted. It does not tell you it should be. Those are not the same question.
What This Means if You Are Buying a Converted Unit
Walk the corridor from the elevator to the unit. Count the seconds you spend without seeing a window. Look at the floor plan and identify which rooms meet the minimum glazing requirement and which exceed it, because the difference between minimum-code daylight and generous daylight is the difference between tolerating your apartment and loving it, and no amount of recessed LED strips disguised as cove lighting will substitute for a window that opens.
Ask the developer what the dark-core space on your floor is being used for, because that space is carrying costs that your rent subsidizes whether it serves you or not. If the answer is "storage units and a gym," calculate whether the storage revenue and the amenity value justify the premium embedded in your rent. If the answer is "we're still figuring that out," you are paying a premium for someone else's unsolved architectural problem.
Conversions can produce extraordinary housing when the bones are right. Pre-1950 office buildings, the ones with narrow floor plates, light wells, operable windows, and ceiling heights that make a studio apartment feel like a loft, convert beautifully. Some of the most desirable residential addresses in lower Manhattan are former offices from the 1920s. But those buildings were designed before air conditioning made deep floor plates possible, and their proportions reflect a relationship between structure and daylight that modern office towers deliberately abandoned.
The algorithm does not know which kind of building you are standing in. It knows the measurements. It scores the factors. It outputs a verdict. The rest is architecture, and architecture is the part that no one has automated yet.
What This Analysis Cannot Verify
Gensler's Conversions+ methodology is proprietary. The 150-plus factors and their weighting have not been published in peer-reviewed form, and no independent audit of the tool's predictive accuracy has appeared in the public record. Bloszies's dark-space measurements derive from San Francisco office buildings and may not generalize to other markets with different building codes, floor plate norms, or seismic requirements. Conversion costs of $500 to $600 per square foot reflect CommercialEdge aggregate data, not project-specific estimates, and vary significantly by market, building condition, and scope of structural intervention. The Brookings financial model assumes specific rent levels, cap rates, and policy environments that differ across the six studied markets. The 90,300-unit pipeline figure from RentCafe counts units in planning or permitting stages, not completed conversions. The 5% trust figure from First American applies to commercial real estate professionals broadly, not specifically to those evaluating conversion feasibility.