A cratering post-pandemic commercial office high-rise market and the ongoing housing affordability crisis have created a complex puzzle to solve. While empty office towers of all vintages litter central business districts, often the only appropriate targets for residential conversion are the oldest on the market: Turn of the (19th) century buildings typically have narrower floorplates and thus more access to windows, natural light, and fresh air.
In an era before ubiquitous air conditioning or even electricity, these smaller floorplates and proximity to operable windows were required to get work done—today, window access is often a legal requirement for apartments. Even so, these century-old high-rises still require dramatic alterations to be converted into apartments, like tearing out the center of the building for a window-lined interior lightwell.
“I don’t think it’s realized how complex these conversion projects are, how much you discover when you open an old building,” says Liz von Goeler, an interior design principal at the design firm Sasaki.
In Chicago, architects and developers are converting office buildings in the moribund LaSalle Street Corridor into 1,600 apartments, with a special focus on subsidized housing; 600 of these will be affordable. In New York City, Mayor Eric Adams is pitching a plan to create 20,000 new apartments with the help of the Office Conversions Accelerator, a group of high-level civic officials meant to organize and marshal resources, expediting the conversion process.
Residential conversions are a priority of the federal government as well, as the Biden Administration has released a Commercial to Residential Federal Resources Guidebook that maps out federal programs that can support this work, with HUD offering additional guidance on how Community Development Block Grants can be used for the reuse of office buildings.
In the second quarter of this year, office vacancies reached a 30-year high of 18.2%, and commercial real estate investment volume has dropped by 64% year-over-year. At the same time, the housing market is short nearly 4 million units, and as of 2021, nearly half of renters are cost-burdened. Beyond the gap in the market that conversions can fill, converting offices to apartments offers critical carbon savings. Early research suggests that renovated structures generate 50–75% less carbon emissions than new construction.
To meet the burgeoning demand for office conversions, von Goeler and Ken Goulding, Sasaki’s director of research and development, have developed a tool that uses a process akin to parametric modeling to analyze high-rise office floor plans and determine ways to divide these spaces into apartments. Sasaki intends to keep the tool in-house, marketing its services to architecture clients and developers.
Called Office Shift Pro, it’s largely a pre-design software for sketching out optimal 2D unit-type mixes in residential conversions, finding the composition that will make a plan financially viable. This can refine the pro-forma gut reaction of how a developer might approach a project with parametric rigor that can splay out and organize all conceivable options, reducing risk.
“They are risky projects,” von Goeler says. “With construction costs high and loan rates high, you have to be very efficient with your layouts to actually make money. These existing buildings have a lot of unique constraints that you have to really think about architecturally.”
Office Shift Pro’s simple interface has users upload a plan-view of an office floorplate and input the desired unit types (studio, one-bedroom, two-bedroom, and so on) and a range of unit sizes for each type, corridor widths, and window access, as well as estimates for rental rates and conversion construction costs. From there, the engine generates thousands of variations across the floorplates, slotting in the units every available way. Users can alter the unit-type mixture (the percentage of studio versus larger apartments) with a set of color-coded sliders, and options are ranked by floorplate efficiency.
These options allow developers to game out different scenarios before spending money on acquisition or construction. Would a series of studio and one-bedroom apartments filled with communal kitchens, lounges, and a fitness center marketed toward young professionals be viable? Or would a building with larger, multibedroom family units with a grocery store and a school on the ground floor be more cost-efficient? It can answer hypotheticals such as, “If we build this out as 80% studios, is that going to make the numbers work?” Goulding says.
Nailing down some of these variables early in the pre-design process can help other pieces slot into place later on. If developers know the programmatic route to profitability early on, they’ll be able to answer, “Where do I need to buy the building, where does the market need to fall in order actually make this a real project?” von Goeler says.
The results of the parametric diffusion are plotted along a two-axis line graph chart, grouping unit mix variations by profitability. All dots adhere to the broad unit mix guidelines users input into the tool, with slight variations. Each dot represents a specific unit mix, and larger dots represent specific unit mixes with more options for where each unit is placed on a given floorplate. In the future, potential enhancements might include integration with daylight modeling and 3D geometry.
It’s a more flexible, less-prescriptive approach than a purely AI and machine learning-driven tool. Office Shift Pro is customizable from job to job and can be tweaked to accommodate retail and amenity spaces as designers and developers see fit. “The tool that we’re delivering is specific to that particular market and that particular building,” Goulding says. Developers can test out, “How much space can we give up within the building to meet the needs of amenity spaces because that’s a direct loss to your income?” von Goeler says.
Researchers elsewhere are investigating other pre-design uses of AI in adaptive reuse projects. Because of adaptive reuse projects’ complex, build-up sites and extensive history, a granular understanding of existing infrastructure is critical here in ways that are deemphasized in new, greenfield construction. Therefore, reality-capture techniques managed by AI can offer a strong advantage. By collecting photos or 3D scans of an existing building to be adaptively reused, and using photogrammetry engines to stitch these images together, AI could be deployed to identify site elements (landscape types, building elements, circulation routes) and tag them with metadata that defines their specific qualities and functions. Pairing reality capture with real-time sensors allows AI to incorporate climate attributes as well as energy load and carbon emission efficiency.
As detailed in the academic journal Sustainability, researchers in China are investigating how AI can be used to adaptively reuse large, shuttered industrial sites in the city of Guangzhou. The group is focusing on developing Generative Adversarial Networks (GAN) which pairs two neural networks and pits them against each other: one that suggests design interventions for a given site and another that sorts through the suggestion and accepts or rejects them, training the original neural network to produce better and more accurate outputs. The GAN technology here is used to identify and classify site elements (roads, pedestrian trails, green areas, different types of buildings, parking lots) and assessments of environmental performance.
Because adaptive reuse projects often adapt high-profile historic buildings with long cultural histories, the Chinese team is also investigating ways to use natural-language models to analyze public statements (like social media posts) about potential reuses and organize them so that designers can easily draw programming ideas from them. These same natural language models could also generate questionnaires about potential reuses and organize the results to create a largely automated public input feedback loop.
As Office Shift Pro demonstrates, much of the potential for AI in adaptive reuse lies in assessing the applicability of existing buildings and dealing with the wide variability in adaptive reuse costs. Radically changing the function of an existing building can be much cheaper than building new, or much more expensive. But as building datasets become more robust and AI gets more sophisticated at sorting through them, these algorithms could formulate meticulous estimates on the cost-effectiveness of adaptive reuse across a wider spread of building types, without the inherent conventional wisdom assumptions that people don’t always look past.
“If you give any of us a floorplan and send us off to put together a set of unit mixes, we’re going to come up with our own biases based on what we think it should be, and it’s really hard to get past,” she says. “This can get past all of those biases very quickly.”
Zach Mortice is an architectural journalist based in Chicago.
Executive insights
AECO
D&M
