Analyzing the AI Hype Cycle; Applied to Automation in the Property Industry

Artificial intelligence and machine learning have been praised as game-changing developments of over the last 5 years, however, many feel disappointed when comparing expectations to real-life application. AI might have missed the mark in the short term, but it has been ‘progressing in the shadows,’ gradually helping to automate tasks and reduce costs. This is certainly the case in the property industry.

But where is AI on the "Hype Cycle," particularly when it comes to Proptech?  According to the graph below, the Gartner general sentiment for the property industry is that AI usage it's overrated and in the hype phase.  

While I cannot comment on a majority of AI subcategories, I believe Gartner overall is too harsh on the property industry. Companies in the property or real estate sector have implemented AI and computer vision implemented successfully and cost-effectively. The appetite for AI tech has also increased in real-estate according to the Deloitte 2020 CRE Outlook which shows a positive future for AI in the industry. We believe that AI as a whole is too vast a subject area to map on such a simplistic hype cycle and in reality, the industry will go through multiple local minima's and maxima's. 

Let's take a deeper dive into how specific companies have cost-effectively implemented AI in the property space.

Deepfinity (Facilities and Property Management):

Our flagship product is called Parcel Tracker; it's an internal parcel tracking and mailroom management software enabled by Computer Vision.

Traditionally parcel management is a manual process at receptions and mailrooms. A parcel would come into a building and be handed off to the staff who would then have to log the details manually and notify the recipient that their mail/parcel had arrived. Doing this a few times is not very time consuming, but buildings now receive 1000s of packages monthly which takes up the majority of the staff's day. Deepfinity has taken this manual process, built advanced Optical Character Recognition Technology and automated the task. What took hours to do only takes minutes. 

With Parcel Tracker, you can quickly scan packages received at the reception/mailroom using a smartphone camera, automatically notify recipients and collect their e-signatures for proof-of-pickup. It works with all couriers and hand-written parcel labels. 

This is an example of deep neural networks being harnessed to do computer vision on a task that could genuinely benefit from automation and save 1000s of hours of work per building. 

Restb.ai (Sales)

RestB provides a Computer Vision API for real estate, focusing on the sales and listing side of the problem. At a core, RestB uses AI to read image contents to tag them, another time-consuming task when dealing with 100s of listings and having to edit and verify them manually. While many such API's exist across the Big Tech Cloud AI solutions, RestB has specialised their system for the real estate market, training the neural networks on different classification domain. 

The most evident specialisation comes in the form of their "Feature Tagging" and "Captioning" API's. They can analyse images and highlight features such as fireplaces, hardwood floors, natural lighting and other features that would typically be highlighted in a real estate listing. This image data can then be combined with location and listing data to auto-generate captions and descriptions for buildings. 

This API has the power to remove a very repetitive human task from listing properties or verifying listings on a platform. What would have previously have required a person to do can now be automated. 

NTrust – (Lease Abstraction)

Ntrust has built a range of data products the most prominent of which is their lease abstraction suite for real estate. Lease abstraction is the process of taking a physical lease contract, digitalising it and extracting information from it. 

NTrust goes above and beyond the traditional lease abstraction providers. It will firstly use deep neural networks it trains weekly to detect and extract the text from scanned documents. After the extracting the text, it will apply a semantics engine to extract and categorise clauses in the document, making them searchable and easily viewable. The AI only get the user about 80% detection and digitalisation, and the remain 20% are done by a human in the background. This ensures that the data is complete.

This automation saves companies 1000 and gives them both a strategic and tactical view of their lease contracts. For example, they can see at a click of a button what percentage of the contacts could invoke force Majeure during COVID and not pay rent. A critical piece of information for planning for cashflow.

Related content: We did an with NTrust VP of Real Estate in EMEA, which dives deep into the tech they employ.

So, what is the recipe for a successful AI solution? 

Each of the AI applications developed above satisfy a number of criteria or questions when producing an effective AI solution: 

Is the task being automated narrow in scope? We have not reached the point of General AI in which machines can deal with complex problems outside of the scope they’re programmed for. Is the task you’re looking to automate specific in nature and repetitive? 

Is the task widespread across the industry? Companies that create automation solutions are working for profit. The markets they tackle must be narrow enough so that they can develop a solution and large enough for them to be a commercial success. 

Will removing the task save money? Some tasks might seem perfect for automation as they appear to take up a lot of time across the industry. However, upon further investigation, these tasks might only take up a few minutes per person per day from a lot of people, not enough for an individual to feel the pain of said tasks. Subsequently, this won’t be an issue they want to solve as the time saved would be marginal. 

These solutions mentioned above currently offer cost-cutting options for property companies or prop-tech providers. This is particularly important to the former as they operate a marginal business and are always looking for ways to grow that margin. For the latter, the automation offered by these AI solutions allows them to increase their scalability factor. The less humans input there is when scaling the platform, the faster the tech provider can grow. 

Aside from the companies mentioned above, AI has been extensively used for fraud detection in payment systems and matching people to properties. These solutions are being developed in the background with marginal improvements that over time have compounded across the industry. 

While currently AI tech has had the most significant impact in reducing operation costs, the real transformation will come when it can offer new revenue streams to Landlords and the property industry. This will be the turning point at which AI and Proptech will become standardised in the industry.