Can AI Replace Property Valuers?

In this digital age where Artificial Intelligence (AI) has dominated numerous industries, real estate is no exception. We now have the technology advanced enough to get property valuation (and very quickly at that!), which has traditionally been done by humans. So the question is: can AI take over property valuers' jobs?

Well before we can answer this question, we must understand AI's role in property valuation compared to humans'. Let's get right into it.

It's pretty straightforward, property valuation is an assessment of a property's value. It is needed for buying and selling, mortgage approvals, taxation, and investment analysis. So, property valuers, or also known as appraisers, would typically assess a property based on various factors. This includes locations, size, condition, market trends, and comparable sales. These professionals use their expertise and knowledge of the market to provide an accurate valuation. But now, AI has stepped in and it changes everything.

But before we jump into how AI has affected the industry, we should first understand the more traditional way to evaluate property.

1. Sales Comparison Approach
   This is one of the most prevalent methods of valuation in Singapore due to the active property market and the availability of comparable sales data. Valuers compare the subject's property to similar properties that have recently been sold, considering factors such as location, size, condition, and amenities.
2. Income Approach
   This method is extensively used for valuing commercial properties and investment properties that generate rental income. It involves calculating the Net Operating Income (NOI) and applying a capitalisation rate to estimate the property's value.
3. Cost Approach
   Though not as commonly used as the Sales Comparison or Income Approaches, the Cost Approach is still used, especially for newer or more unique properties. This method calculates the value based on the cost to replace or reproduce the property, minus any depreciation.
4. Residual Method
   This method is often used for valuing land sale or developments that have been marked for en bloc. It estimates the land value by subtracting the development costs (including profit margins) from the expected completed value of the property. Given Singapore's active real estate development scene, this method is quite relevant.

AI-powered tools can process vast amounts of data from various sources like historical sales data, market trends, economic indicators, and property characteristics. And because machine learning algorithms are very good at identifying patterns, it's able to predict property values with remarkable accuracy, or at least that's how it's supposed to be. In real life, it's not always that simple.

Take Zillow, a major real estate marketplace in the United States, as an example. They introduced an AI-driven feature called Zestimate, which claims to "use a sophisticated neural network-based model that incorporates data from county and tax assessor records and direct feeds from hundreds of multiple listing services and brokerages". Basically, they throw in a bunch of big words to convince users that their model is reliable. But, despite having a median error rate of only 2.4% - 7.9%, Zestimate can only be as accurate as the data on their listings. So, if a listing is outdated, or, if a seller inputs the incorrect property details, then the Zestimate of that listing will ultimately be unreliable. And the moment the listing is updated, the Zestimate can change drastically. Personally, I would say that it's quite ironic how the technology that's supposed to help minimise human errors comes with its own set of errors. But to be fair, Zillow did state that Zestimate is not a proper appraisal.

That being said, many establishments in Singapore have also recognized AI's potential in property valuation. Well-known companies like PropertyGuru, 99.co, and SRX Property have all dipped their toes in automated property valuation. And it seems like this model has found more success in Singapore, especially with HDB flats, for several reasons.

Firstly, unlike Zillow where listings can be updated by just anyone, most real estate marketplaces in Singapore rely on professional input. Typically, property agents are the ones responsible for both posting and taking down listings on these platforms, making sure that the details are all correct. This means that the data fed into AI models is more reliable, resulting in more precise valuations.

And because development of HDB flats are managed by the government, they have standardised designs and sizes. This uniformity reduces the margin of error significantly, making automated valuation much more reliable and accurate.

What's more, transactional data is highly available. Even the government maintains comprehensive and up-to-date records of property transactions, a crucial database for accurate valuations. AI tools can tap into this database to enhance their predictions.

For example, a team from TechinAsia used URA's record of apartment and condo open listings. Using this along with other data sources like geospatial data and satellite imagery, their machine learning model was able to predict property prices with 87% accuracy. And mind you, this was back in 2019. So imagine the improvements they've made since then.

For all the property valuers biting their nails reading through this article, here's the part you've been waiting for. Yes, your service is very much still needed.

Although AI-powered valuation tools do quite well in Singapore, human valuers remain irreplaceable. While AI can analyse quantitative data with high precision, humans bring unique skills, experience, and contextual understanding to the table, elements that AI cannot fully replicate.

For one, human valuers understand the nuances that can impact property values. This includes location, overall condition of the property, surrounding amenities, school districts, upcoming developments, economic shifts, and market disruptions. Not to mention subjective factors like architectural or interior design, renovations done, materials used, historical significance, superstitious beliefs, and other buyer's sentiment. Human valuers can easily adapt to these ever-changing factors and adjust their valuations accordingly, whereas AI may struggle to keep pace.

Achieving such a level of expertise takes years of experience and cannot be entirely captured by AI algorithms, something that the TechinAsia team acknowledged. It was actually this issue that prompted them to use nontraditional sources to create something different - a model that sets them apart from other automated valuations that only rely on historical data and broad market trends. But even their model can't compete with professional valuers.

Moreover, in many jurisdictions including Singapore, legal and regulatory frameworks mandate that property valuations be conducted or verified by licensed professionals. This ensures that valuations adhere to established standards and practices, providing a layer of accountability that AI systems alone cannot offer. Human valuers are trained to provide assessments that are both legally compliant and professionally sound.

AI may not have taken over the property market, but as it continues to evolve, its role in property valuation is set to expand further. And although it's unlikely that AI will ever replace humans, they will undoubtedly be a useful tool for professional valuers. Having AI to automate data collection and analysis can reduce human error and increase efficiency. But ultimately, it still needs human inputs and oversight to yield reliable results, much like how our agents use PropNex Investment Suite. Even though this tool has been very helpful in providing data and market trends, we rely on our agents' ground experience, expertise and wisdom to make a proper analysis.

As the real estate market continues to embrace technological advancements, the collaboration between AI and human expertise will definitely shape the future of property valuation.

Views expressed in this article belong to the writer(s) and do not reflect PropNex's position.