The future of artificial intelligence remains a question — will it emerge as an ally in addressing the housing crisis, or will it be a force that accelerates the concerns?
As America faces increasing pressure to provide secure and economically viable housing, AI and other emerging technologies are increasingly cited as tools that could transform the affordable housing industry. Yet uncertainty remains about how best to apply them effectively.
While the sector evolves to meet growing demand, it’s essential to examine how innovation might advance its mission and improve outcomes for both residents and operators.
Before fully embracing these tools, we must ask: What are the potential pitfalls? And more fundamentally, what can AI realistically contribute to this space? Can it be meaningfully leveraged to support those developing and maintaining affordable housing and, more importantly, to improve the lives of the residents these homes are intended to serve?
To evaluate its real-world impact, we must first ensure we share a clear understanding of what artificial intelligence really is.
So, What is Artificial Intelligence?
While there are official definitions of what exactly constitutes artificial intelligence, the term has become so widely adopted that it describes vastly different systems. Some software can transcribe audio and video recordings, while others can generate images using online databases, and others convert various file types into a unified format for streamlined collaboration, and the list keeps growing.
Information for this graphic is primarily sourced from NASA: ‘What is Artificial Intelligence?’
While there can be some overlap in these systems, the use cases for each piece of software can be so disparate that simply calling each one ‘AI’ is akin to going to the zoo and saying that you saw some animals and offered no further clarification.
“It is possible that generative AI will bring some of the work efficiencies and breakthroughs that venture capitalists and big tech have been hyping for years, but we must grapple with the fact that the main ways that people encounter generative AI is as brute force internet pollution,” wrote Jason Koebler in a recent article on 404 Media.
The noise, both the hype surrounding it and the concerns about its use, continue to grow in volume, and cutting through that to understand what it means and how it can be used effectively remains a major question. It’s clear that this type of technology will remain in some shape or form for the foreseeable future, but how can it be used to tackle unwieldy data sets, streamline the steps throughout the building and housing application process, all while tackling the ethical and environmental concerns that it presents?
How Can AI Impact Affordable Housing Developments?
AI software and the systems surrounding them are an exceptionally sharp double-edged sword. With tools as powerful as those now being developed and deployed, their success depends largely on the intention and care behind their use.
One currently accessible and potentially massive time-saving use of GPTs — generative pre-trained transformers, a type of artificial neural network that is trained on large data sets to generate responses to prompts and queries — is to significantly shortcut the amount of time it takes to get answers to specific and granular questions related to government policy. Uncovering the needed answer can take considerable time and effort and often includes sifting through endless pages of text.
“We are developing custom GPTs for different government agencies,” said Michael Willard, a Strategic Operations Associate at MRK Partners. “So, for the California Tax Credit Allocation Committee (CTCAC), you can upload all of their publicly available PDFs, the information within surrounding deadlines and submission nuances, their different links, and you can make a custom model that has a base knowledge of what CTCAC is.”
The same can be done for the Missouri Housing Development Commission (MHDC), or for the U.S. Department of Housing and Urban Development (HUD), or for any of the other governmental agencies that require strict compliance in order to ensure affordable housing developments can be built.
Understanding how these models work and how to adjust what they’re pulling from is an important step in cutting down on the hallucinations — generated responses riddled with false or misleading information — that are increasingly common from these systems. Another key step MRK takes very seriously is fact checking these responses against both internal knowledge and the documents it’s pulling from in order to confirm validity.
“Adding these PDFs…cuts down on potential hallucinations,” said Willard. “If you ask a general GPT a specific HUD question, it’s going to pull from either the internet or what it’s trained on. And that can be incorrect or inaccurate given the highly specific requirements, but if it has in its pretrained memory, like in the custom GPT, this additional documentation, it’s going to reference that specifically.”
It is vitally important to prevent any hallucinations from making their way into final documents. Not only would it cause cascading problems throughout the process, but there are also potential legal ramifications involved that need to be considered. Because of this, MRK restricts the use of these models to spaces where any output can be thoroughly fact-checked and closely scrutinizes all output to ensure that it meets the highest standards.
Managing the paperwork, asset management, and other bureaucratic steps are a significant part of any successful affordable housing development. They often involve multiple agencies, organizations, and departments to balance funding, protocols, permissions, approvals, and countless other requirements that need to be completed in order to even think about putting a shovel in the ground.
Similarly, there is considerable thought put into managing and protecting this data if and when it is fed into these custom models. Lackadaisical implementation of these custom models, in addition to being of legal concern, could also expose the information used to both fraud and cyber-crime. Protecting the data and information used, and potentially restricting its use to ensure security, is a key step ever-present throughout this process. It’s important to note that private personal information of residents is never uploaded to any model for any reason — these systems are used to assist with the bureaucratic side of the process, and MRK will always prioritize the safety and protection of its residents, employees, and any privileged internal information.
Actually, building the affordable housing our communities need is the cathartic end of a long and sometimes tedious process, and digging that first hole is just one success in a string of many that need to happen before new residents can move in.
“Another custom GPT we made is for these things called handoff memos,” said Willard. “There are a lot of different relationships and different requirements involved in setting up a deal, and every deal is individualized. So what we’re doing is training GPTs on the different reporting requirements, the different loan documents, the different memos, and at the end, you’ll have a proprietary custom GPT that you can ask specific questions to that should be able to generate responses for what you’re specifically looking for.”
Security Concerns and Understanding AI Implementation
Blindly relying on this technology is a clear and present concern for both MRK Partners and anyone looking to push the adoption of these technologies. There are a growing number of lawsuits against many of the major tech players in this space, and ethical concerns over how their data sets were gathered in the first place remain a shadow hanging over these systems.
When creating a custom GPT, understanding these risks and acting accordingly is paramount to ensuring that compromising or private information is not introduced into these systems. “You don’t want specific documents like letters of intent, or some of the other agreements publicly available,” said Willard. “In creating the MRK workspace, it removes entirely that training ability, and it’s encrypted end-to-end. You have to leverage both pushing the advancement of AI while also keeping security in mind.”
In the future, there is the potential to create non-internet-based GPTs through internal servers that counter the risk of secure and sensitive documents being introduced to a public sphere. In the meantime, limiting what can be uploaded to publicly available information and documentation is a simple way to ensure that you’re protecting yourself and others involved from a potentially harmful information breach.
Understanding the potential pitfalls and dangers of any new technology is a key step in its intentional implementation, and the potential benefits of its effective use are staggering. MRK Partners, and other companies that manage properties and projects across the country, need to be able to sift through terabytes worth of data, documentation, and files to ensure that each location is both meeting local and federal requirements, but is also meeting the standards set by the company itself.
The Value of Maintaining the Human Element
As AI becomes more efficient at streamlining different tasks, the questions around what it means for employment become louder. A recent World Economic Forum survey indicates that 41% of employers intend to downsize their workforce because of AI automation in the next half decade, with certain roles taking significantly harder hits than others. Whether the intention is to cut costs or reinvest in other areas, the potential threat to millions of people’s livelihoods is ever-present.
On top of those growing concerns, questions around the value of human interaction continue to blossom. When it comes to something as personal and valuable as housing, is it really worth cutting out that human connection and replacing it with a computer interface?
Some areas, particularly those inclined to jump on the latest technological advancements like San Francisco, are implementing zero-person touch point leasing strategies. In some apartment units, every step of the process, from applying to taking a tour to the application and review process to receiving the keys to a new apartment, is handled without ever coming in contact with a person. While the concept may excite some, it can also prove to be an insurmountable roadblock to others.
There is a baseline of technological competence required to navigate these types of systems — not only that, but there’s also the requirement of a baseline of equipment. While most people have a smartphone, not everyone does. While most people may know how to manage the process through different apps and systems, not everyone can. For some, the process can appear seamless and a glimpse into the future. For others, it’s a barrier to access that could leave them without a home if there are no other options.
“A fully automated, zero human interaction leasing process integrates technology across every stage of the resident journey, from initial inquiry to move-in,” said Willard. “In a conventional multifamily setting, this process can be largely digital. However, in the affordable housing sector, a hybrid approach is essential — leveraging automation where possible while maintaining accessibility for residents who may have limited access to or comfort with technology.”
This automation can begin as early in the process as the listing. Certain platforms use this technology to syndicate listings from major services, and then automated leasing assistants can respond to questions in real-time and help prospective renters schedule either self-guided or assisted tours, limiting the need for manual coordination from on-site staff.
Online application portals can also assist in collecting applicant information, including identity verification and documentation processing. This technology can enhance the initial required steps to confirm certifications, eligibility, and ensure compliance standards are met.
These technologies can support every stage of the process, from resident screenings and leasing generation to payment management and coordinating move-in logistics, all within a streamlined digital framework.
“Ultimately, these systems do not eliminate the need for human interaction — they refocus it,” said Willard. “By automating routine tasks like scheduling, follow-ups, and data entry, leasing staff are freed to provide higher-quality, personalized support where it matters most: helping residents complete critical steps, resolving exceptions, and maintaining meaningful engagement throughout the tenancy. These human touch points ensure equity without undermining the efficiency gains of automation.”
Balancing the technological benefits while mitigating the harms is a vital part of adopting and implementing this technology and understanding how to marry the value of AI with the necessity of human connection is what may very well set companies apart from their counterparts.
Community remains a core part of what MRK wants to bring along with its affordable housing opportunities.
“We don’t want the zero-touch point leasing,” said Willard. “We want the managers and staff at our properties to be able to interact with our residents more. Every time you’re doing something that could be semi-automated, you’re taking time away from focusing on the human component. Instead of using this technology to transfer responsibilities, you’re letting them focus on better uses of their time.”
Where AI Implementation Can Represent a Danger to Affordable Housing
Systemic oppression and racial inequality have been at the core of housing issues throughout history, and advancements in AI technologies both have the potential to start addressing these issues and exacerbate them. Because this type of technology functions on the data and information it is given, the algorithms in play have the potential to follow along the same dated principles that have plagued housing equality for generations.
“The principles are not only dated, but the principles are even shaped or formed by the lingering effects of past discriminatory policies,” said Dr. Michael Akinwumi, the Chief Responsible AI Officer for the National Fair Housing Alliance in an interview with the Pennsylvania Human Relations Commission. “There are a lot of examples, especially in housing, where many of these discriminatory effects are still continuing to shape the algorithms that we’re using. For example, any time we look at Home Mortgage Disclosure Act data, we see the patterns, and it’s clear to us that this is just another redlining but in digital form.”
When designing an algorithm intended to positively affect affordable housing rates, or housing processes in general, blindly feeding information pulled from past and historical documents to create it will simply perpetuate the same problems that those processes faced in the past. Creating and implementing these systems requires dedicated attention and intention in order to avoid those pitfalls and protect those who need that protection the most.
“AI does risk increasing inequality,” said Phebe Vayanos, an Associate Professor of Industrial & Systems Engineering and Computer Science at the University of Southern California in a TED Talk addressing the potential of designing AI to advance social equality. “Many communities and groups, those who need it the most, have been excluded from the benefits that these technologies can bring, and existing tools cannot simply be applied in these contexts… We should all be directing more of our time, resources and energy to designing AI to support the most vulnerable, marginalized, and underserved communities.”
Understanding policy and why certain policies were implemented is a key step in ensuring an equitable system can be put in place. Not only that, relying solely on an algorithm to make a decision and entirely removing human oversight runs the risk of spiralling out into new and aggravated problems. Housing is a fundamental and necessary part of everyone’s life, and its care cannot be placed entirely in the digital hands of AI or other similar technologies.
Navigating the Crossroads of AI and Affordable Housing
The intersection of artificial intelligence and affordable housing is not a matter of if, but how. As the sector contends with mounting pressures from regulatory complexity to operational efficiency and equitable access, AI offers undeniable promise. Yet that promise is inseparable from a web of risks, responsibilities, and social considerations.
What we know for sure is that AI is a tool, one with immense power to streamline inefficiencies and free up human capacity for deeper engagement with residents. But wielded without intention, oversight, or an equity-focused lens, it could just as easily amplify some of the systemic issues it hopes to resolve.
In this pivotal moment, the question is not whether AI is friend or foe, but how we will choose to shape its role. The answer lies not in the technology itself, but in the values we bring to its design and deployment.
Tackling all aspects related to the future of affordable housing is a key part of MRK Partners’ mission. Learn more about our thoughts on how insurance costs are affecting the industry, or check out our brand new TV feature on Empowered, hosted by Meg Ryan to hear about the company’s mission to push the affordable housing industry to new heights.