I spent a large part of my career in the mortgage business – long before computer vision was a thing. I was a VP at one of the largest nationwide retail lenders (Great Western), a Fannie Mae Director both in public affairs and later as head of the Seattle Partnership office, and finally, a mortgage loan officer for over eight years with the largest lenders (Countrywide, Wells Fargo and Chase) in the country at the time.
These combined experiences give me a unique vantage point of looking at this industry from a 35,000 foot level and the surface as a former boots-on-the-ground originator.
Having also worked with innovative FinTech firms, I’ve become particularly jaded by the years of bravado touting one-week or one-day closings. For a couple of decades, dozens of VC-funded new players said they would disrupt and/or revolutionize the mortgage industry. I’m still waiting.
But every decade or so, a new technology that earns the title “revolutionary” by every measure emerges. AI, machine learning, and computer vision – which are interrelated – all fit this bill.
The AI race
Now the AI race is on. But what most people in real estate don’t realize is the “AI race” started a very long time ago. My first encounter with AI was at Fannie Mae in 1995. One of my Public Affairs Office teammates and I met with one of the three people in our tech center down the street on Wisconsin Avenue. We were looking at what tech innovations Fannie was working on for possible PR stories.
We discovered a small group was working on something called “Artificial Intelligence,” which, at the time, I thought was pure science fiction. We could not write about it as we quickly learned their research was not to be shared publicly.
Fast forward to an announcement this week from real estate’s computer vision leader, Restb.ai. Restb.ai has already blanketed the MLS world by embedding its tech inside every major MLS platform provider.
Their newest announcement this week is intended to modernize the appraisal industry.
Restb.ai has what may be an insurmountable lead in computer vision in real estate: they’ve been around longer – since 2015 – so they have more hours clocked to better their tech. Startups can’t catch up in AI the same way they may be able to with other technologies, as more time developing AI tools equate to greater learning and more improvement. Restb.ai pushes the benefits, but they also know the dangers of computer vision and have built their tech with the required guardrails (like Fair Housing-compliant auto-generated descriptions and data security).
Now Restb.ai is expanding its tech to address appraisal modernization, an area where Fannie Mae has been helping to lead the way. For example, most recently, Fannie made a significant shift: it now evaluates photos for condition and has rejected appraisals when comparables are used without the appropriate adjustments.
To fear or not to fear AI?
As with any substantial change, many appraisers are fighting AI’s entry into their space. They fear this movement will be their demise as AVMs continue to improve as the tech gets smarter.
Restb.ai sees things differently. They see computer vision as empowering appraisers, not replacing them. Launching a new Valuation Product Suite, Restb.ai is offering the appraisal industry a slew of advanced tech tools for Comparable Properties, Data Collection, Form Pre-Population/Validation, and, perhaps most importantly, Restb.ai’s proprietary Appraisal Complexity Score.
Power of computer vision
Computer vision also might be vital in addressing racial bias in AVMs. According to the AEI Housing Center, comparing the inside condition of a home could be a big deal in helping to understand differences in properties.
Responding to the PAVE Working Group recommendations on AVM bias, which you can read here, AEI Housing Center wrote:
“We also found that home interior quality can be a significant factor in the difference between an AVM and a home’s sale price and that for a small number of properties with extreme conditions, that difference was significant.”
That’s what makes computer vision so interesting is the multitude of potential uses: from appraisals to real estate property listings, it unlocks data we never could collect as fast or completely as before and can help automate some of the most mundane tasks that can streamline the workflow for agents and appraisers alike.
Here’s the thing: this is only the beginning of the use of computer vision in real estate as we are in the very early days, and even Restb.ai is just scratching the surface. We are witnessing one of the most exciting times ever for our industry as we watch the fusion of technology and real estate finally being actualized.
You can read more about the Restb.ai valuation announcement here.
