Maurice Grassau, CEO of Architrave, cuts through the buzzwords and hype to explain what machine learning is and what its use is in real estate.
Define machine learning in one sentence
Machine learning is a subfield of artificial intelligence, which broadly seeks to make machines that perform complex tasks by imitating intelligent human behaviour.
Without getting too technical, how does machine learning work?
There are two distinct different ways machine learning can operate. One method uses rule-based algorithms, and the other uses a self-learning model.
Take a digital game of chess. In rule-based machine learning, you explain how each piece can move and then you play against a computer that follows those rules. In the self-learning approach, you give the algorithms enough data so that the machine learns the rules itself.
What does machine learning enable you to do that you couldn’t otherwise do (or would take too long)?
We use machine learning to standardise and classify documents. That might not sound sexy, but when you consider that having all of your documents and data clean, standardised and clearly classified on one platform can help you act and move much quicker in, for example, a recession, machine learning can be an essential asset.
At Architrave we use a hybrid machine learning model. That means we use machine learning to help our employees work much faster when classifying data. While a human can process 20 documents per hour each day, our algorithms can process around 50 in the same period – with human help.
The reason we use a hybrid model is quality assurance. Combining machine learning with a human to ensure accuracy adds a level of quality control to every document we process.
What types of proptech platforms use machine learning?
While it’s not a proptech, Apple Mail is a great example of how people use machine learning every day without realising it.
If you receive an email via Apple Mail, you’ll occasionally see a square box around the sender’s signature. Apple Mail’s machine learning recognises the text elements of a signature as an address. It then compares it to signatures within your address book to see if your contact details match. If not, it asks you if you want to add a new contact. This is machine learning at work.
PriceHubble is one example in proptech. The Swiss startup uses machine learning to predict the right price point for residential real estate. For example, imagine that you own thousands of units. Using machine learning, PriceHubble will process this data and tell you how much rent you should ask for in each unit.
Archilyse is another great example. The company’s machine learning reads 2D floor plans and uses them to create full 3D BIM architectural models of what buildings will look like when constructed. This allows stakeholders to assess the value of potential properties.
A lot of tech platforms say they use machine learning. What questions should people ask to figure out whether there’s actual machine learning taking place?
One important aspect of machine learning is training data. So, one question someone could ask is how much training data a tech platform has and where they get it from. If they don’t know, that’s a red flag.
If everyone uses machine learning, is it actually interesting and innovative, or is it a standard feature you should expect from proptech?
What’s important is to ask whether using machine learning in your business creates real value for your clients. If a proptech creates value using a tech setup that doesn’t include machine learning, then they have a good business. Machine learning is just one of the tools in the arsenal depending on how you create value. Some people will use it, some people won’t. It’s not necessarily a badge of honour if you do.
Maurice Grassau is CEO of document and data management business Architrave