Category: Tech

• Can AI democratize property tax appeals? Doesn’t look like it.

  Property tax appeals are a routine part of owning commercial real estate. And there are lots of people who will help you with an appeal. One common way that these consultants charge for their services is as a percentage of achieved savings. This creates a low-commitment scenario for landlords and a strong incentive for the consultant to perform and get paid. The same is true for individual households, who also engage in property tax appeals. However, research shows that some groups are far less likely to appeal:

  Lower-income homeowners are substantially less likely to file an appeal. Even holding home values constant, Hispanic and Black homeowners appeal at lower rates than White homeowners, while less-educated homeowners appeal at lower rates than more-educated homeowners. These disparities are often attributed, at least in part, to differences in knowledge of the tax system, confidence in navigating the appeals process, and the ability to afford a human agent (Doerner and Ihlanfeldt, 2015).

  But what if more tools and support were provided?

  Here’s an interesting study. To test this, the researchers recruited a sample of 645 households in Dallas County, Texas and gave each of them a website providing personalized property tax information and instructions for how to file an appeal. However, half of the households were assigned to a “treatment” version of the website that included a Claude-powered AI chatbot, which was there ready to answer tax-related questions and provide personalized guidance.

  What they generally found was that homeowners liked the chatbot. 78% of those who had access to it initiated a conversation. The researchers also found that using the chatbot meaningfully increased the probability of filing a property tax appeal. The baseline increased from 41.4% to 50.5% (a 22% increase!). Importantly though, the increase in appeal filing was smaller among less-educated homeowners, those with lower-valued properties, and those from racial or ethnic minority groups. Overall chatbot take-up wasn’t all that different, but the translation into action was.

  This is interesting because it provides “suggestive evidence” that simply providing access to AI isn’t enough. In fact, it had the opposite effect in this study. Rather than reduce inequity, it exacerbated it by increasing filing rates among the already more advantaged.

  2026.09.29

• From hours billed to value created

  Large law firms typically operate on a structure known as a pyramid or leverage model. The basic idea is that at the top there is a small group of equity partners who are expected to make it rain and bring in big clients, but who don’t actually do most of the work.

  Beneath them is a middle layer of senior people who manage the day-to-day, and at the very bottom is an army of juniors who do most of the actual grunt work.

  Because the base is wide and the juniors are getting billed at hourly rates that far exceed their fixed salaries, the partners at the top get the benefit of the excess funds flowing upward. It is a model that relies on juniors working long hours, most people burning out and leaving, and a relative few becoming partners.

  But as we know, AI disrupts everything. Interestingly enough, Wall Street firms are reportedly going to their big law firms and saying, “Hey, now that AI is empowering you to work fewer hours and be way more efficient, we’d like you to tell us exactly how many hours you’re saving and reduce your fees accordingly.”

  According to this recent FT article, the broad expectation is that the cost per transaction is going to come down significantly and that legal firms will need to adopt different business models in the short term.

  I don’t think this necessarily means that legal firms will become less profitable, but it certainly encourages people to move away from “this is how many hours I worked” to “this is all the value that I created for you.”

  2026.09.15

• We are now in a post-camera world

  

  This is a wall of 20th-century filmmaking equipment captured at the Fondation Jérôme Seydoux-Pathé. For background, Jérôme Seydoux is a French businessman and movie producer (and the grandfather of French actress Léa Seydoux). Pathé is the second-oldest operating film company in the world. It was founded by the Pathé brothers in the late 19th century. (For architecture fans, the Fondation itself was designed by Renzo Piano. Note: Full architecture tours only happen once a week.)

  

  

  Seeing this taxonomy of equipment and reading about its market adoption reminded me that technologies change, but human impulses tend to remain remarkably static. Film cameras at the beginning of the 20th century were made for industry. They were expensive, bulky, and required teams of people. But as you’d expect, by the middle of the century, these devices became more affordable and suitable for individual use. This gave rise to amateur filmmakers and eventually, you could say, proto-influencers.

  The introduction of smartphones made creation even easier and social media platforms completely changed the game of distribution, turning content creation into an even bigger business. But in the end, the human desire to capture, create, and share remains the same. Today, however, the game is fundamentally changing. AI eliminates the need for handheld devices and physical models and locations. Now you can just prompt. The possibilities are endless.

  As I was walking around the Fondation and reading about how it took a year to cast the right female lead for L’Amant (Jane March was eventually found in the suburbs of London), I couldn’t help but think about how AI completely collapses this process. Today, there’s no need to look; you just need to describe exactly what you want. Invariably, this is going to feel like some kind of loss to purists. Isn’t this artificial?

  We are now in a post-camera world. But what will never change is our desire to tell stories.

  2026.09.05

• Cynics sound smart but optimists build the future

  There’s an old saying that “cynics sound smart, but optimists move the world.” And indeed, studies show that we typically perceive pessimism as a sign of intelligence. On some levels, this makes sense because it requires intellect to understand something, and then expose its vulnerabilities.

  But at the same time, pure cynicism can be inherently passive. It’s an easy armour to wear, one that protects the user from (1) taking risk, (2) being proven wrong, and (3) doing the hard work to determine whether something might actually work. Importantly, cynicism generally creates zero new value. It may protect you from a bad decision, but it also blocks you from the really great ones. Value creation demands action.

  Now, blind optimism is not the answer. Risk management is essential. In fact, some of my real estate developer colleagues describe the business purely in these terms: “I manage risk for a living.”

  The right answer is a kind of rational optimism. Because the greatest value creation comes when you believe something is true or possible, even if it has never been done before. If someone has already proven the thesis, it doesn’t require the same degree of optimism; it only requires research.

  If you’re looking for evidence of this in practice, consider that the most entrepreneurial cities in the world have robust capital markets that are inherently tolerant of risk. Venture capital, for example, is rational optimism scaled across an asset class. It operates on a power-law distribution. The assumption is that most investments will be a failure, but a select few will more than make up for it.

  This is an important feature because it rewards bold ideas that can change the world. But for all of this to work, you need a healthy dose of institutional optimism. Without it, these bets go unmade and the future gets created by someone else. Let’s not ignore this key ingredient as we build our cities.

  2026.08.28

• How AI is transforming architectural workflows

  I recently came across an online post criticizing a Canadian Tire ad that was very clearly created by AI. The text was illegible (most AI models aren’t excellent at text). The model’s face was plasticky. And everyone was piling on, saying things along the lines of: “Shame on Canadian Tire. They should have hired a Canadian model, photographer, and graphic designer.”

  While I can appreciate where this is all coming from, the reality is that the cat is out of the bag. As I’ve said before: What is real anymore? It doesn’t matter. We all have to adapt. I have, for example, noticed a dramatic change in the architectural workflow on our projects. We are now seeing an infinite number of visualizations throughout the design process, and that is making it a lot easier to iterate and refine ideas. So much for the quick blue foam models.

  On some projects, where we have a collaborative working relationship with the design team, we now send AI renderings back and forth in group chats: “I’ve changed the material on these soffits and modified the brick coursing. What do you think?” None of it is perfect. But neither were the physical models and other tools that we used to rely on to test ideas.

  If anything, it’s overwhelming in the best possible way. The creative possibilities are endless.

  2026.08.26

• Data centres will soon claim 15% of total US grid capacity

  Here are some recent data centre figures to help put things into perspective. By 2035, BloombergNEF now expects there to be 194 GW of data centres online in the US. This is an upward revision of 83%! The reason for this revision is that “the announced data center pipeline in the US has grown by a further 101 GW since December 2025.”

  But here’s another interesting thing. BNEF also estimates that the shipment of AI chips between 2023-2033 could total as much as 325 GW of data centre capacity in a theoretical world with no other constraints. Of this figure, they expect 207 GW to go to the US.

  However, they also identify a shortfall of 63 GW between this expected supply of AI chips and what is actually forecasted to get built due to grid constraints. So in other words, the chip market may be overshooting what can actually be physically deployed (though there are nuances to consider here).

  Regardless of what ultimately ends up happening, we’re talking about enormous numbers. The recent clean energy investment here in Canada is expected to generate an additional 14 GW of clean, renewable power. It goes to show just how many of these big announcements we are going to need for Canada to become a data centre superpower.

  At the end of 2025, the entire utility-scale electricity-generation capacity of the US was 1,280,799 MW, or about 1,280 GW. So when we’re talking about an install base of 194 GW by 2035, we’re talking about ~15% of the country’s current grid capacity.

  2026.08.25

• The great bifurcation: Machine-centred vs. human-centred real estate

  “When you fall in love again, you don’t think about the ex very much.” —Travis Kalanick, co-founder of Uber, talking about how he feels about Uber today

  You’d be forgiven for thinking that Travis Kalanick, co-founder of Uber, is back. Since leaving Uber heartbroken in 2017, he has been relatively quiet. But in his words, “I’ve been working my ass off the whole time. I just haven’t been talking about it.” Clearly. Last month, he announced that his startup Atoms had just raised $1.7 billion in equity. The round was led by venture firm a16z.

  Atoms is an audacious business and I would encourage you to check out their website and read their vision. Travis describes his life’s work as “digitizing the physical world” and Atoms’ mission is “physical automation to transform industry and move the world.”

  What that means — and what’s fascinating for us real estate people — is that he’s creating a physical AI company that uses single-purpose machines (as opposed to general-purpose humanoids like what Tesla has been promising for a number of years) to automate specialized tasks such as food preparation, resource extraction, and transport.

  However, all of these “atoms” need to be stored and processed somewhere, and so Travis views land and real estate as a critical resource in the world of physical AI. He goes on to argue that real estate development is a “dramatically underappreciated” ingredient. Everyone in tech has been too focused on software. And he means it. I took a look at their careers section and they have dozens of real estate-related job postings across the world.

  The hope with all of this is a new Golden Age:

  “Everything in our world, in our cities, in our civilization – look around you – is mined or grown – manufactured and moved. The next Golden Age will be upon us when the means of growing, mining, manufacturing and moving physical things becomes fully divorced from human labor.”

  When you consider this vision and our current data centre boom, it’s not hard to imagine the potential impacts on our built environment. Our cities are bifurcating into two broad forms of real estate: machine-centred and human-centred real estate.

  The machine-centred assets might service our online lives or the physical world, as is the case with Atoms, but they won’t necessarily need to engage with their broader environment or have things like windows, climate control, and even lights. They are “black boxes” for machines.

  Human-centred assets are, on the other hand, the sort of spaces that we talk about on this blog: walkable, human-scaled communities where people genuinely want to live, work and play. This, in my view, is the divide we are headed toward and it’s going to mean a reconfiguration of our urban areas and a repricing of real estate assets.

  Intuitively, I think this will only enhance the value of human-centred assets. We are going to want what is scarce: authentic human connections, beautiful architecture, and differentiated people-first experiences.

  2026.08.15

• Who actually wins the autonomous vehicle future?

  There was once a time when Uber was heavily investing in autonomous vehicles and rides. Then that stopped in an effort to become profitable. But now, out of necessity, it’s back. Uber recently announced that the company will invest “$10 billion of capital over the coming years to bring AVs to market at scale.”

  Uber’s belief is that the self-driving car industry won’t be dominated by just one company. Instead, it will be a mixed ecosystem where some companies build their own cars and ride-hailing apps, and others plug into existing networks. And in this world, it is these networks — like Uber — that will become the most valuable piece of the puzzle.

  Perhaps.

  Whether they believe this or not, Uber basically has to say this because they are the platform aggregator. They do not build or operate any self-driving vehicles of their own — at least not anymore. These parts of their business were sold off.

  Autonomous vehicles are a wicked technical challenge. The edge cases are extraordinarily difficult to solve. If only a select few companies solve them, and solve them well, then there’s an argument to be made that they will capture the majority of the market.

  But if AV technology becomes commoditized and it ends up being all about the brand and who can aggregate demand the best, well then, Uber could very well be right. They already have global scale.

  My view is that Waymo has such a commanding lead when it comes to AVs that they have the leverage in the short and medium term. There are also zero switching costs for ride-hailing customers. I’ll use whatever is better. So if I’m Uber, I’m worried about Waymo, which is presumably why they are starting to have differing opinions.

  2026.08.13

• Canada is uniquely suited to become a global hub for data centres

  Kevin Yin frames the data centre debate eloquently in this recent Globe and Mail article: “The backlash against data centres is understandable. But the answer is not to block construction. It is to design a better bargain.”

  Here are some things we can say about data centres right now:

  • They create relatively few long-term direct jobs (though innovation spillover does exist when they cluster).
  • They are power hungry.
  • They are a core physical engine of the new global economy.
  • They play to Canada’s physical and economic strengths as an energy superpower.

  Meaning, Canada is uniquely suited to become a global hub for data centres and AI infrastructure. Canada has abundant energy, naturally lower cooling costs, geopolitical stability, and deep institutional capital.

  The challenge is that we need to manage the negative externalities. No household, for instance, wants to pay higher energy bills.

  We also need to make sure that there are long-term economic benefits for the country, and that we’re not repeating the old habit of “exporting” our resources to other countries so that they can innovate. That would be a bad deal.

  The world is going to continue to need a lot more data centres, and Canada is in a unique position to lead and have control over its own destiny. Kevin offers a few ideas for what a better bargain might look like, if you’d like to have a read.

  2026.08.06

• What happens when Uber doesn’t have to pay its drivers?

  This is an interesting video by Phil Andrews of Maxinomics talking about the economics of Uber and what it could stand to gain from autonomous vehicles.

  As part of this, he touches on the exclusive partnership that was announced between Uber and Waymo back in September. That was a big deal.

  I find this topic fascinating because it’s hard to imagine it not reshaping the landscape of our cities. And it continues to get more real by the day.

  2024.11.18