The adoption of artificial intelligence (AI) is expanding exponentially throughout daily life, aiding both how we live and work. As the large language model and predictive analytical tools that dominate AI applications are being refined, users are gaining confidence in the results of AI-assisted tasks.
Setting aside concerns about AI’s potential for negatively disrupting the economy or replacing human beings, it is already possible to see how that technology can automate many (or most) of the menial tasks associated with a wide spectrum of industries and professions. Chat bots seamlessly handle most of the inquiries about products and services in nearly every industry. Research tools can “scrape” billions of documents and follow a simple series of prompts to produce a well-written report in minutes that would have taken days of manual effort. Predictive tools can assimilate mountains of data about transactions, patient symptoms, or football plays to more accurately forecast sales, diagnose a disease, or suggest a game plan to attack the Steelers’ next opponent.
The housing market is as ripe for improvements through AI adoption as any other industry. The processes by which sellers’ market or buyers shop for a home have dozens of repetitive, time-consuming tasks. Critical decisions, like setting the right price for the home, evaluating the credit worthiness of a borrower, or estimating the cost of construction, are built on a foundation of data from previous sales or projects.
Adoption of AI has been relatively enthusiastic by the real estate industry and more grudging by the construction industry. Buying or selling a home requires research and processing of loads of information and documents, many of which are tediously repetitive. The more successful realtors are often those that gain an information edge. It is logical to use AI tools that enhance and accelerate any information edge. Homebuilding, by contrast, is a more conservative industry. Roughly half of every dollar that a builder spends is for labor at the job site. There are informational and technological edges to be gained in homebuilding, but there are also greater risks in construction that make the businesses involved in residential construction less likely to be early adopters.
These broad industry characteristics are influencing AI adoption thus far. According to the National Association of Homebuilders (NAHB), the value of AI in residential construction was $5 billion in 2025, a fraction of the $905.2 billion total for new construction last year. By contrast, AI-driven real estate transactions totaled $226 billion in 2024, according to PwC and the National Association of Realtors (NAR), and is expected to reach $734 billion in 2028. That would be roughly 40 percent of the total of all residential transactions.
The application of AI in the housing market will almost certainly be widespread by the end of this decade, probably performing tasks that we are not anticipating today. Looking forward from four years away, there are already numerous activities that AI is currently improving in the buying, selling, and building of homes.
AI in Homebuilding
The 2025 NAHB survey of AI adoption among its homebuilder members got a bit more into the weeds about how the technology was being utilized. While one in five builders claimed to be working with one of the large language model applications, most of the tasks being tried were tied to creative activities or market research. Only five percent of the builders responded that they were using AI for business operations, such as estimating, design, or automated equipment operation.
Lower rates of adoption by homebuilders tracks with the industry’s reputation for resistance to change, but perhaps the slower adoption is the result of the technology’s perceived irrelevance to the two major problems plaguing residential construction: too few lots and too few skilled workers. But those that are utilizing AI are clear that it is a tool that saves a great deal of time. That is a resource that builders can never have enough of.
One of the most visible applications of AI in home construction is in design. Generative AI tools allow architects and builders to rapidly explore design options based on defined parameters, such as lot dimensions, budget constraints, local building codes, and energy efficiency targets, producing optimized layouts in a fraction of the time required by conventional methods, including computer-aided design (CAD) and building information modeling (BIM). Autodesk, the leading developer of CAD and BIM software, studied recent housing projects and documented savings of half the cost and time when its AI-augmented design tools were utilized.
AI-enhanced design tools can also harness the massive amount of information needed to track project-specific code or energy requirements, test the effects of differing HVAC systems or window selections on the efficiency of the building, or process countless variations in structural or material choices to produce designs that homeowners can easily compare to make critical decisions in real time.
Estimating and project management are two other areas where builders are using AI to enhance performance. The enormous level of detail, combined with enormous amounts of project-specific information, involved in estimating and managing the construction of a new home make those tasks ripe for technology that can learn from each experience. AI further augments the benefits of managing large amounts of data by marrying its communications capabilities. For example, estimators can utilize AI to automatically and quickly verify real-time costs and availability across the entire scope of work just ahead of a client meeting. Project managers can use AI to follow up action items from meetings automatically, and flag those items that remain unresolved before the next meeting.
AI is also being married to hardware to address some of the physical challenges of home construction. Some of these, like robotic workers or self-operating equipment, are solutions of the future. Others, like the use of aerial or fixed site monitoring systems, are improving safety and productivity now.
In the Pittsburgh market, most builders report that AI is being tested, if not integrated into their operations.
Matt Rost, regional market manager at NVR, Inc., the region’s largest homebuilder, reports that Ryan Homes and Heartland Custom Homes have only recently begun using AI applications, which are too new to report much in the way of results.
“We haven’t fully adopted it yet, though we’re evaluating where it makes sense. Not quite there yet,” says Liam Brennan, vice president at Infinity Custom Homes.
Those builders that are more fully integrating AI into their businesses are effusive in their praise of the benefits.
“We are utilizing and encouraging employees to utilize it in as much as we can. It is a strong tool for research, clerical, and design ideas. It’s not a major part of our business, but we are utilizing it more and more each day. I don’t have a way to measure it yet, but it saves so much time,” says Jeff Costa, president of Costa Custom Homebuilders “We took all of our brochures and marketing materials, all our plans and specs, all our documents and uploaded them into AI to create a ‘Costa bot’ that knows everything about all our homes.”
Costa gave an example of an AI solution that he would never have expected prior to adopting it. One of his clients, for whom Costa Custom Homebuilders completed a new home a dozen years earlier, called with a major underground plumbing problem. The problem resulted from the use of a copper-to-copper connection that Costa does not regularly use underground. By searching its trove of documents, and relevant documents on the Internet, the AI application found a short-lived requirement by Pennsylvania American Water that was in place when the home was built. The homeowner was able to resolve the problem and Costa’s good reputation with a client remained intact.
“My brother thought he remembered something like that happening in 2010, but it would have taken so much time and trouble to try to find that with Google on our own,” Costa says.
“The most interesting use is for rendering,” he continues. “If a customer takes a picture of their kitchen, goes to the Internet and finds a table they like, they can ask AI to show their kitchen with that table. They can do the same with a type of trim, or flooring, or whatever. I would have to pay a designer, and it could take a week or two to come up with renderings that we can now come up with in 20 or 30 seconds. Our clients can come up with an exterior elevation and change colors to see what it would look like.”
“We are using it as a tool for efficiency, research, rendering, document review/compare, among other things,” says Paul Scarmazzi, CEO of Scarmazzi Homes. “The president of our company is very progressive. He’s very focused on what works. He won’t follow shiny objects but as he engages in AI he’s trying to work through how it will be adopted. Some of our people are saying they’re sticking with what they know works and others say they will try it because they get it. Some in the field have adopted it, some are reluctant.”
Scarmazzi hits upon an interesting point. In an industry in which performance is the ultimate measuring stick – did the work get done on time for what was estimated – how much effort should an employer put into forcing adoption? While it seems obvious that AI is saving lots of time and effort, it is difficult to document the savings at this point in the development of the technology.
“I can’t honestly say we have key performance indicators or anything to measure it. Because we have strong leadership, it’s something that’s being embedded and people are being challenged to use AI,” Scarmazzi explains. “We’re using it by department where it’s relevant and we’re challenging them to use it, not dictating it. There aren’t measurable outcomes because it’s not defined yet but it’s a rotation back to the people who are using it to say how are you becoming more efficient, how are you using it?”
“Our land department, for example, is using it to track the correspondence and responses with the planning department,” he continues. “We put a project in the development process and a year later, we’re wondering what changed. AI has the ability to overlay the changes as red lines on prints to understand what is revised. We’re using AI in land meetings to process the output of the notes and assign tasks. I’m not sure how to measure that yet, but it’s hard not to see the effectiveness of it.”
AI in Lending
Banks and other lending institutions have the opportunity to use AI along two main channels to improve performance. One channel is administration, using AI to increase efficiency in how the lender communicates internally and externally, manages data, or performs customer service, for example. The other is specific to its core business of lending, that of processing and underwriting loans.
Both applications of AI would improve a bank’s efficiency and effectiveness, but the potential application of AI to loan underwriting could positively mitigate the lender’s biggest risk: making bad loans. AI can analyze an enormous universe of past loans to determine what factors have the greatest impact on a borrower’s ability to re-pay or to default on a loan.
Banking regulations do not outright prohibit the use of AI in loan underwriting; however, residential lenders are heavily scrutinized by several federal agencies to ensure fair lending practices. Agencies such as the Consumer Finance Protection Bureau (CFPB), the Federal Reserve Bank, and Office of Comptroller of Currency (OCC) regulate and audit federally-chartered lenders, including banks, to ensure that AI systems comply with existing fair lending, consumer protection, and civil rights laws to prevent algorithmic discrimination or “digital redlining.”
To use AI in underwriting without violating regulations, institutions must meet several core requirements:
- Fair lending and bias mitigation: Algorithms must be constantly tested to ensure they do not create a disparate impact on protected classes.
- Explainability: If an AI denies a loan, lenders cannot provide a generic denial reason. Under the CFPB’s Circular 2023-03, lenders must provide the specific, accurate factors the model considered.
- Model risk management: Federal regulators require lenders to independently validate their underwriting models, document their data sources, and maintain robust human oversight.
- State laws: State-specific rules (such as the Colorado AI Act) are beginning to impose explicit governance, audit, and impact assessment requirements on AI-driven credit decisions.
“I’m far from the expert on AI and how it’s going to integrate in our world, but I hear and see things,” says Mike Henry, senior vice president at Dollar Bank. “We aren’t currently using AI in our lending process. From an appraisal standpoint it feels like it’s already happening with automated evaluation models. From what I hear, underwriting will be the first area where we can offload some of the work of underwriters into a faster analysis based on the information we have, but I honestly don’t know what that will look like.”
It is the appraisal field that seems ripest for AI adoption. One of the earliest and most mature AI applications in residential real estate is automated property valuation. Automated valuation dodels (AVMs) use machine learning to analyze historical sales records, property characteristics, neighborhood demographics, school ratings, crime statistics, and current market conditions to produce estimates of value in a few minutes.
While consumer home search portals like Zillow and Redfin have well-earned reputations for questionable price estimates, which are not AVMs, platforms like HouseCanary have elevated the accuracy of these models. HouseCanary’s AVM reports a median absolute percentage error of just 2.8 percent, a level of precision that rivals traditional appraisals. Accurate AVMs reduce the time and cost of appraisals, eliminate certain forms of human bias that have historically infected manual appraisals, and enable quick pricing decisions. The enhanced speed of appraising then enhances the speed of loan processing.
AVMs have significant limitations. They depend heavily on data quality, and where comparable sales data is sparse accuracy degrades. There are also growing concerns that AVMs trained on historical data may perpetuate patterns of undervaluation in neighborhoods that have historically been underserved by mortgage lending, thereby perpetuating unfair lending practices. But, as more appraisers utilize AI, including the use of AVMs in the preparation of estimates of value, the models will gain reliability. That represents an existential threat for appraisers.
“We create documents, appraisal report documents. AI has been very beneficial in the creation of those documents. It’s great at synthesizing and producing data in a digestible format,” Connor King, certified commercial real estate appraiser at Nicklas King McConahy. “AI is an incredible time saver.”
King adopted AI relatively quickly, opting to try to learn how it could augment his efforts and become informed about the role it could play in changing his profession. He exercises caution about how he shares information, choosing to use versions of AI that do not require that he allows the data to help the model learn. King is not naïve about the possibility that AVMs could replace the appraiser.
“I do think AI could eventually be pretty threatening. The biggest barrier right now is that much of the information you need to produce a credible appraisal is very specific. A firm like ours, with over 25 years’ experience in the direct market, has lots of market-specific knowledge. That’s what separates us from something that has general knowledge,” he says. “But the bigger the collection of data, the more accurate and, therefore, more credible the opinion of value will be. For now, our reports are still coming from a certified professional. Whoever signs the report is still responsible for the judgement of value and AI can’t do that.”
By the end of the decade, it is possible – if not likely – that AI will be trusted to do the widest scope of tasks that humans perform with expertise, things like disease diagnosis, contract negotiation, or forecasting. At present those adopting AI applications cite benefits that relate to the elimination or reduction of nonessential tasks. While that means that administrative and support jobs will be eliminated, it also means that businesses and professionals will have more time to perform for their customers.
“I believe it will be Skynet someday but, in the meantime, there is still the need for human intervention. You’ll still need to know the answers to the test,” jokes Scarmazzi. “It offers a tremendous opportunity for us to buy back time.” NH