Does AI Remove a Real Estate Investor’s Abritrage?

This was originally produced as a podcast, you can listen below or here. But for those of you who prefer to read, please you enjoy the below.

Does the use of Artificial Intelligence in real estate development and investing diminish a company’s (or an individual’s) competitive edge?

This question arises most frequently when people see generative design tools space planning a site. They see tools like Delve or Parallelo where you just give it a site plan and ask it to design a building.

Pretty much straight away it can come up with numerous options. If it’s trained properly it should be the most efficient layout for whatever the requirement for example: 50% site cover, 50% 2-beds, 25% 3-beds and the rest 1-beds.

When you see this type of kit in action it’s compelling.

If you’re a developer or an investor looking at numerous sites, it’s exciting to see you can get a good level of accuracy, close to instantly compared to waiting for feasibility work for weeks.

However, the following series of questions then unfold:

OK, so if everyone looking at this site is using this tool, do we all build the same thing? And does that mean we all bid the same price? How do I differentiate myself?!

It’s quite a daunting question.

And it doesn’t just apply to design.

It applies to the market analysis.

For example, data companies like Placemake.io collate and draw out conclusions from a vast variety of data sources that no single human could compute previously. Previously we may have had a few pieces of digestible evidence: comparables, footfall data, rental growth forecasts.

Now, companies like Placemake can overlay thousands of ther data points: market data, current supply, current infrastructure, current demographics, proposed developments, proposed infrastructure, the proposed changing demographics, the yearly pattern footfall not just averages or spot checks, weather data and so.

Then their algorithms can spot patterns and make projections that no individual human could and it will say right, on this site you should build this mix of apartment types and your rental growth will be this.

Zoom out again and think about these things on a strategic or portfolio level, if every fund manager is asking some AI  “what can I do to minimise my risk and maximise my returns what different sector, what cities should I be investing in” Will all the major funds all suddenly have the same investment strategy.

Will we have so much data, transparency, universal market knowledge and then the ability to process and analyse that data, that all opportunities for arbitrage, to “outperform” the market simply go away.

Zooming back into single buildings: If AI can identify the "best" use of a site, do we all end up following the same blueprint?

I think it’s a valid concern but I think it will only apply to some scenarios.

MODELLING THE BUILT ENVIRONMENT

Our world, our built environment seemingly has an infinite number of variables.

Think about the butterfly effect. When one tiny change in one variable has a knock on effect that effects the whole world. This is really why, despite having access to the same data, the outcome of development projects can vary significantly due to the infinite variables at play.

When we take it back to this idea of generative AI creating the same designs, you could just look at the example of a family deciding to build an extension on the back of their house. Take the standard London Victorian / Edwardian 3 to 4 bedroom house. People love to put big bifold doors out to the garden and dormers in the roof for a loft extension. You can imagine that an AI tool could pretty quickly produce that standard design, however, every single family has their own patterns, ways of living, style preferences, budget meaning that there is not one single design that becomes the “right” answer.

Again, this is  evident when you think about the complex balance of stakeholder interests in large regeneration projects. When you have a site with numerous buildings and use classes the diversity of potential arrangements is vast. I don’t believe that an algorithm could come up with a layout that was categorically the highest and best use for all stakeholders, not least because, the way those large spaces are redeveloped then changes the way people flow around them so it wouldn’t just be predicting the ultimate layout for today but trying to model those changes over time and again, I think that takes us back to the land of close to infinite variables.

Now, there is one area where I believe AI could effectively take away the developers edge.

And that is on single buildings or single sites that are being developed for some kind of commercial use (and I’m including in that build to rent or sell). Here, the constraints are often clearer. Say if you’re in a conservation area with very clear planning context, building lines and materials, it could be clear you’re not going to stick a steel 10 story tower in the middle of Georgian brick terrace.

On a small and constrained site there might be a very clear way to maximise light, ease of access and services within these constraints. And then beyond the physical imagine in the residential context, could be an area with a clear demographic, ie strong need for 2 bed, 2 bath flats, broadly this price range.

WHAT ARE THE IMPLICATIONS?

Initially there could be an arbitrage for first tech adopters

In the residential example, using AI to generate these designs, get nice logical servicing placement, rendering surrounds, assuming you can get to planning quicker, get to technical design quicker with fewer mistakes, reruns. This could certainly get an arbitrage, if only of a few months, which in the context of a few year build is helpful but perhaps not a complete shift.

But then what happens in a few years time imagine a world where that site is being sold as a residential development and everyone is using these tools. They generate a perfectly efficient building, include great knowledge of the market, so the AI “knows” that this demographic love X size of balconies. In theory this should put all purchasers on an equal footing. And really it should be better for the customer. The developer who wins is the one who can build a good quality building, quicker to a design with fewer flaws.

But what then are the levers for one purchaser to pay more or less than another? Strategy and operations, post development become a more important lever. Is this a build to sell or build to rent or build to short short stay? How well is it managed once it’s built.

And of course developers with lower cost of finance and lower return requirements may well be the ones who win.

It’s worth saying that this isn’t too far off where we are already. Already operations are becoming a key way that investors differentiate themselves and find alternative routes to return.

So I think for the sites with single uses and many constraints, those will be the first where we see the opportunities for arbitrage being reduced by using generative AI tools for the layouts, for the servicing structure and so on but again the variety in the potential for operational strategy is still so great, that there will not be zero arbitrage.

COMPARISONS IN OTHER INDUSTRIES

An interesting comparison is to look at algorithmic trading in the stock market which has obviously been happening to some degree for decades. This takes out human error and human biases and it allows for products such as index funds which many people like for their slow steady tracking of the market. But obviously at the moment algorithmic trading still happens in a complex world with seemingly infinite variables and so they don’t capture black swan events. And yet there is both a market for algorithmic trading and for human-directed investment strategies.

WHAT HAPPENS IF AI CAN FORSEE ALL SEEMINGLY INFINITE POSSIBLE OUTCOMES?

If we get to a point where AI is intelligent enough to make accurate predictions of everything because it can forsee every possible iteration and predict precisely what is going to happen, at that point, we’re in a very different world and future and none us will be worrying at that point about whether we can get some arbitrage on one investment strategy over another because either we’ll be living in a world of pure abundance so it won’t matter or we’ll be approaching the end of humanity so it won’t matter.

REALISTICALLY, WHERE WILL WE END UP?

So in summary, will we end up in a world where everyone has the same investment strategy and there is no potential for arbitrage?

There will be efficiencies, faster iteration of designs, more exploration of different design options, more resilient testing pre purchase, quicker and deeper market analysis and demographic predictions. And certainly less of the spotting an opportunity that no one else has seen. I think AI will deepen this but it’s already a trend we’ve seen particularly as markets become more transparent. Given that our built environment is non-linear and there are infinite scenarios, the ability to operate those assets well and respond to the black swans, the unforeseen events those will continue to, they already are, and they will continue to be, key areas where investors and developers are able to differentiate themselves.