The application of AI in real estate investment management
16. February 2021 | Drooms Global
Artificial intelligence (AI) has become synonymous with innovation. Offering unique capabilities, AI processes are not just automating mundane tasks in real estate but are built to self-improve over time benefiting asset managers, portfolio managers, brokers, sellers, and investors alike.
Before we delve into one of the most popular, real time applications of AI in the sector however, we take a closer look at the technology itself, its subsets, and limitations.
Encompassing a branch of computer science, artificial intelligence is often applied to the development of smart machines adept at performing tasks typically carried out by humans.
In the 1950s, AI broadly referred to computers being able to carry out complex tasks such as playing chess and answering mathematical calculations to the same degree or faster than humans. Nowadays AI has broadened in scope and is applied to a variety of industries including healthcare and e-commerce.
AI can be subdivided based on different criteria. For example, AI algorithms differ based on what representation is used internally to model knowledge about concepts and their relations. In symbolic representations, one symbol represents one real-world concept. Statistical models represent concepts as vectors of numbers, usually derived from analysis of large data sets. Neural networks encode concepts as sequences of activations of sets of neurons in the network.
AI can also be categorised based on its application area:
- Natural language processing deals with analysing textual data and includes tasks such as translating a document or extracting key information from a document e.g., the tenant name from a lease contract.
- Image analysis is concerned with analysing images (e.g., a page from a document) for tasks such as classifying the image (e.g., is it a photo or a map?), detecting objects in the image (e.g., does it contain a signature?), or locating objects in an image (e.g. where on the page is the signature?).
- Generating predictions and suggestions is typically based on analysing data, e.g., user interactions in log files. This information can help to address challenges such as grouping users together based on their actions (e.g., novice, intermediate, or expert user), or predicting or suggesting what their next action should be. This is similar to product recommendations based on past buying behaviour.
At Drooms, we are using all three approaches (i.e., symbolic, statistical and neural network models) analysing both text and images as well as generating suggestions.
AI and machine learning are used almost interchangeably but there is a clear distinction between the two, as illustrated in the diagram below.
AI is an area of computer science concerned with creating technology to mimick human behavior to make machines more intelligent. Machine Learning is a subset of AI, that uses statistical learning approaches to produce self-learning systems that improve and correct their outputs over time. Deep Learning is a subset of Machine Learning which is typically based on artificial neural networks. These artificial networks are inspired by the biological neural networks and by how a human brain processes information. Recent innovations and improvements in AI-based technology such as self-driving cars and major improvements in machine translation technology are all based on deep learning.
Using state-of-the-art approaches such as deep learning yields better performance (e.g., higher accuracy), but it typically requires a substantial amount of data training and the produced results can be difficult to explain or to correct.
A recent example of exaggerated expectations involves GPT-3, a huge language model trained on billions of sentences. Example applications include automatically filling in missing values in a spreadsheet or generating software code based on a natural language description. To a non-expert, this might look like magic. What is usually not mentioned is that this model will either fill in values in the spreadsheet which were part of the training data or completely invent new numbers, which may or may not be correct. For the example of generating code automatically, the output is often not syntactically correct (“does not compile”) and would still require manual corrections.
With the advent of self-driving cars and other autonomous solutions, questions surrounding accountability have resurfaced. For example, who is responsible when a self-driving car is involved in an accident? Similar questions arise when AI empowers fully automatic decision-making. At Drooms, we have adopted the strategy of always keeping a “human in the loop” to confirm or reject suggestions made by a machine, so that the user is still always in full control.
Nevertheless, AI will continue to permeate into technical solutions such as data rooms to help automate time-consuming or repetitive tasks.
Drooms has built artificial intelligence into its data room offering. Developed 100% in-house, AI is applied to daily workflows with the goal to profoundly improve repetitive yet fundamental and time sensitive tasks.
Revolutionising the work of real estate professionals by streamlining and improving efficiencies in due diligence, sales and the investment management processes, Drooms launched a series of award-winning features including but not limited to:
Document Translation: to enable users to work in their preferred language and facilitate real-time secure analysis of all the facts pertaining to an international asset sale.
Auto Allocation: to automate the allocation of documents to their respective index points. The algorithm takes all actions into account, learning from individual behaviors and patterns and refining the document's future index allocations over time.
Auto Naming: to save time lost reviewing scanned documents and renaming them in the data room environment by automatically categorising documents and generating and assigning file names based on their actual content.
For more information on AI at Drooms click below.