Briefing Note #1
Intro to the Briefing Note
The purpose of this note is to help individuals and organisations in the real estate world navigate AI and its potential impact on the industry.
Everything will be grounded in the question “and what does this mean for the real estate industry”.
This briefing note will provide you with insights, news, and tools in a format that you can read or listen to every fortnight. We will look to lessons from other industries and progress in our industry. Questions that you should be asking yourself and your company in your work.
The ultimate purpose is to leave you feeling more confident in navigating the world of AI in Real Estate.
This note is produced by Hattie Walker-Arnott, who trained as a chartered surveyor, worked in commercial investment & development, and then moved to a tech company. She has focussed on AI since mid-2022 alongside working as a development manager for regeneration projects.
AI in Real Estate So Far In 2024
Where Are We Now in May 2024 - 2023 saw the rise of AI in the public knowledge, therefore naturally the question arose: how can AI be used for real estate investors and developers?
As we stand, there are no tools I have tried that make me say “this is essential for you to remain competitive, as a business, today.”
Instead, there are emerging real estate-specific tools that are starting to demonstrate the potential of AI. Additionally, there are a number of more general business tools like Microsoft Copilot, Zapier AI integrations and obviously Chat GPT which if embraced can significantly speed up your day-to-day business operations. There are practical steps that individuals and companies can start taking to adopt AI and understand its potential and current fallibilities. We’ll share 5 of these steps over the coming weeks.
Why is AI the right tech for Real Estate? Real estate has been notoriously difficult to digitise. Each asset is physically, financially & legally unique and they change over time. This makes it difficult for traditional computer programming to capture all elements of a building. AI should change this. AI doesn’t need a human to program every single input and possible output. Instead, an AI model can learn from what it has seen before and make its own conclusions. We will be able to use AI models to look at the existing data such as maps, surveys, images, videos, lidar scans, thermal scans and so on and build AI models to spot patterns and reach conclusions. I’m hopeful that AI will help us to quantify why, for example, living rooms with high ceilings feel better. At the moment we call that gut feel, instinct, or experience but AI should help us articulate this more precisely.
AI as a Fundamental Technology - AI is a fundamental technology. It’s more like electricity or the internet than a specific company or app like Uber or Instagram. It is a tool which will make all sorts of other things possible. This fundamental nature means solutions will arise that seem like magic/science fiction today. And we cannot accurately predict the tech that will emerge in 1, 5, or 10 years. Think about it this way: when the internet launched in the 80s, we couldn’t predict that the same technology would create the potential for people to have whole careers as “influencers” or “social media advisers”. With AI we can confidently say there will be some mega changes and improvements but exactly what they are will unfold over time.
AI Ethics: With social media, we came to regulation too late. People got trolled, became depressed, anorexic but eventually the social media companies started putting policies and systems in place to deal with it. We don’t believe have the luxury of time to wait to see what goes wrong with AI before putting rules and regulations into place. Compared with social media, the potential influence of AI is on a greater scale. I’m sure by now you’ve seen deep fake videos, the deep fake phone calls and can imagine the influence these could have on our society and our economies. Thinking deeply about ethics has not been on the day-to-day agenda for many professionals in recent decades beyond things like the Bribery Act. With AI we will all have to become conversant with the ethical questions and implications. For example, are you comfortable using image and language generators if you don’t know if they’ve been trained on images they have the right to use? What if you don’t need to hire so many graduates once you’ve adopted tools? With AI we are trying to emulate human intelligence to speed up processes that we humans would rather not do. That, in itself, sounds noble. But who are the humans designing the technology that is emulating the intelligence? And how can they create technology to “support humans” when every human has a different opinion of what that looks like.
Glossary
GPT - Generative Pre-Trained Transformer
A GPT is a type of machine learning model designed for understanding and generating human-like text. It's pre-trained on vast amounts of text data to learn grammar, facts, and reasoning, and then fine-tuned for specific tasks. “Transformer” refers to how the machine learning model works. Transformers use an “attention mechanism”. This enables the model to "pay attention" to the most important parts of a sentence and how it relates to other parts. This improves its understanding and generation of language.
GPT coined by OpenAI in 2018 (link)
Transformer coined by Google in 2017 (link)
Quote of the Fortnight:
“Good lawyers will become more like legal therapists - creating trust with clients and helping solve their problems, rather than simply writing briefs and doing research.”
Thoughts of the Fortnight
Someone should create a method to measure biodiversity on sites using accoustic monitoring
In a year or two saying your business is “AI-powered” will sounds as ridiculous as saying your business is “internet-powered” or “email-powered” today.
Key News Stories of the Fortnight
OpenAI Release ChatGPT 4o (link) - OpenAI are upgrading their free members to ChatGPT 4 and introducing a new, faster version for paid subscribers: Chat GPT 4o, “O” for “Omni”. ChatGPT 4o can interpret and respond with information from images/voice/code meaning it can provide deeper and more natural responses. My initial thoughts are that this will create a great inspection copilot to share pictures and ask opinions whilst on site. Will play some more and share other experiences.
Costar Buys Matterport (link) - This is interesting because Costar has rent/value data and Matterport has spatial data. Combined, they could deliver insights into the true value drivers of space and design. However, in my experience, Costar data is not always entirely accurate. If AI models are trained on this inaccurate data, then the results will not be so useful (and this applies to any model trained on existing but potentially inaccurate data).
Amazon, Alphabet & Microsoft post strong AI-driven results, Meta bucks the trend (link,link,link) - Amazon, Microsoft and Alphabet have posted strong earnings with increased revenues of 13%, 17% and 15% respectively, all mainly attributed to their AI efforts. Meta reported weaker sales forecasts and higher spending (on AI). This caused an immediate share price drop of around 10%. Since this has recovered to a drop of around 5%. Some say that the investor attitude is due to a lack of confidence in Zuckerburg’s spending aspirations given his previous track record with the Metaverse.
JLL CEO Highlights AI in Investor Demand (link) - Christian Ulbrich states data centres are the “hottest asset class at the moment, many people are trying to get into them”; “when you believe in AI, the demand for data centres will only go up”; “the shortage is usually the grid” and that data centres are “a play for the next 5-10 years”. All of these are fair comments. Certainly in the short term, I agree, demand for data centres will outstrip supply. If you’ve got new space with electricity supply you’re in the money. In the longer term, I’m not so confident. Data Centres obsolesce quickly. Koomey’s Law states that electrical efficiency of computation doubles every year and a half. Yes, we’ll need more computing power for AI but that doesn’t directly translate into data centre demand. If you’re already in the sector you’re golden. If you’re not, don’t believe it will be a surefire winner.
AI Fundamentals - Machine Learning
Exploring frequently used terms
Machine Learning is a branch of artificial intelligence (AI) focused on building computer systems that learn from data, rather than following explicitly programmed instructions. You can think of it like a graduate in a valuation rotation.
Training: The graduate surveyor starts by learning from their more experienced colleagues, reading past valuation reports and going on inspections and asking questions. This is similar to training a machine learning model using historical data.
Prediction: Now, when the graduate is given a new property to value, they enter it into Argus, looking at other examples. They will adjust according to their inspection, comparable research and conversations with the team. This is like the model making predictions based on learned patterns.
Feedback: If the initial valuation is too high or too low or uses the wrong comparables, the experienced valuers provide feedback, and the graduate surveyor refines their understanding. This is akin to improving the model's accuracy through new data and feedback.
Specialization: Currently the graduate is specialising in valuation. They might move on to leasing or investment next. In their new rotation, they will use the same principles (rents, yields, locations, build costs, interest rates) but they will spend their days doing different things (running cash flows, drafting leasing brochures, writing valuation reports). In machine learning, these are like specialized models (algorithms) designed for specific tasks.
Summary: Machine learning is like training a graduate surveyor who becomes increasingly skilled at valuing properties over time by learning from past data and refining their assessments through feedback.
Disclaimer: This content is for informational and educational purposes only. It does not constitute financial, investment, legal or IT advice.