The AI property tech debate nobody's having: why Greystar thinks architects are the answer, not the problem
The world's largest residential developer isn't using AI to replace architects. It's using it — via Giraffe — to put them back in charge.
There's a version of the AI-in-property story that goes like this: generative design tools get good enough that you no longer need architects for early-stage feasibility; algorithms spit out optimised apartment plans; developers cut costs and move faster. It's a compelling narrative. It's also, according to one of the most data-driven residential development businesses on the planet, largely wrong.
Gary McLuskey is Global Design Director at Greystar — a business operating in 17 countries, managing over two million residents across more than a million apartments globally, and delivering somewhere between 20,000 and 25,000 new apartments every year. For context, the largest developer in London builds around 2,000.
He's also been working with Giraffe for several years, and recently joined its board of directors.
His take on AI and architectural design is worth paying attention to — not because it's contrarian, but because it comes with an unusual amount of evidence behind it.
"It's all terrible"
Greystar has spent serious time exploring generative design. Generative buildings, generative apartment plans, algorithmic optimisation across all of it. McLuskey is blunt about what they found.
"I'll be honest, it's all terrible when you really kind of understand what works."
That's not a dismissal of AI broadly. It's a specific critique of a particular application — using generative algorithms to replace design judgment — and it comes from a business that has surveyed thousands of residents, built a proprietary design assessment tool, and has more data on what makes residential communities successful than almost any organisation in the world.
What Greystar has found, consistently, is that what residents respond to isn't optimisation. It's uniqueness. It's quality of experience. It's the difference between a building that feels designed for people and one that feels designed for a model. An Airbnb collaboration — Airbnb monitors every interaction residents have with every listing, and shares those insights with design partners — reinforced the same finding: there's no single winning formula. What wins is distinctiveness.
Algorithmic generation, at least in its current form, is not great at distinctiveness.
The actual problem: architects have been pushed out
The more interesting argument McLuskey makes isn't about AI at all. It's about who's been controlling development projects for the last 25 years — and the answer, increasingly, isn't architects.
"The architect barely registers. It's probably uniquely worse in the UK than any other country. In the US, architects have a lot more control. But it's just been an erosion of the architect's role."
He describes an expanded cast of consultants, cost managers, project managers, and specialists who have gradually absorbed responsibilities that architects once held. The consequences are real and measurable. Cost control — arguably the most important lever in any development — has largely passed out of architects' hands. Program management has followed. Architects have drifted toward being design consultants rather than project leaders, which means the people with the deepest understanding of what makes a building good are also the people with the least influence over the decisions that determine whether it gets built, at what cost, and on what timeline.
McLuskey's argument is that AI represents an opportunity to reverse this — but only if architects are looking in the right direction. The risk is that they focus on generative design tools (the AI applications that are most obviously aimed at architects) while the real transformation is happening elsewhere: in structural engineering agents, services engineering agents, planning consultant agents. AI is automating the specialist roles that displaced architects in the first place.
"We've now got a structural engineer agent that's doing stuff for us. A services engineer agent. A planning consultant agent. These are the things that, as architects, we need to take control of."
The implication is significant. If AI can handle a meaningful portion of what quantity surveyors, structural engineers, and planning consultants currently do, and architects are the ones orchestrating those tools, the architect returns to the centre of the project. If developers use those same tools to cut out architects further, the profession loses even more ground.
How Greystar is actually using Giraffe
Against this backdrop, Greystar's approach to Giraffe makes a particular kind of sense. They're not using it as a generative design tool. They're using it as a platform — one that connects design decisions directly to financial outcomes in real time.
The workflow starts before a site is found. Greystar operates on what McLuskey calls an "advanced design" model: you understand exactly what product you want to build for a given market before you go looking for land. That means having apartment typologies, block configurations, and building forms resolved to a level of detail that most developers only reach after they've acquired a site. By the time you're running underwrite analysis, you're not guessing at what the building will cost — you already know, because you've already designed it.
Giraffe sits at the centre of this. As the architect draws, they see construction costs, rent values, and residual land value updating in real time. The model connects to a wider dataset — costs, sales data, market rents — so that every design decision is made with full financial transparency. A team in Germany can open the same model that the UK team is working on and see the massing, the underwrite, the construction costs, all of it.
The result, McLuskey says, is that they can take any site and have a full set of construction drawings and a costed design within a week.
"That's incredibly powerful. And we know the costs and we know the timescales."
What changes when architects have cost visibility
The less obvious benefit is what happens to design conversations when financial transparency is built into the process from the start.
"Rather than spending tons of time talking about 'this apartment doesn't work and you've got to fix that,' we're spending all our time talking about the urban spaces and the experience. What kind of features should we have in the building that people would really like? What kind of urban space would work with the community?"
The low-value problem-solving — the iterative fixing of apartment layouts that don't work against the controls, the back-and-forth on compliance — compresses dramatically when the design is developed against a well-understood brief with real-time feedback. The high-value conversations, about place, community, and the things residents actually care about, expand to fill the space.
This is, in McLuskey's framing, where technology and AI are actually leading the industry — "into a world that's a lot deeper in terms of thought." Not toward a future where fewer people are needed to deliver a project, but toward a future where the people who remain are spending their time on genuinely hard problems.
The data behind the design decisions
Greystar's design group is built around research. The business runs a continuous resident survey programme, maintains a design assessment tool that scores projects against resident preference data, and produces detailed guidance documents for architects working on every product type across every market.
Some of what that research turns up is counterintuitive. A recent survey of over a thousand residents found that 29% would prefer north-facing apartments — a finding that directly contradicts the planning policy rationale for orientating buildings away from north aspects.
McLuskey's point isn't that north-facing apartments are good. It's that blanket policy rules — "a blunt instrument that doesn't respond to the real world," as he puts it — are a poor substitute for understanding what people actually want.
Greystar's best performing assets, the ones where residents stay longest and communities are most stable, are mid- and low-rise developments. Not towers. Not the high-density vertical approach that dominates the policy conversation about housing supply. Buildings where residents can see the street. Buildings where the relationship between private space and public realm is human in scale.
"People would rather live where they can see the street than in a tower looking across at another tower."
This matters for developers and investors because it's a direct challenge to the assumption that density and height are the same thing. Barcelona, not Sydney or Melbourne, is McLuskey's reference point for how to hit urban density targets while building in a way that residents actually choose to live in long-term.
Platform, not algorithm
The distinction McLuskey draws between Giraffe as a platform and generative design as an algorithm is worth sitting with.
"Giraffe is more of a platform than it is a sort of algorithm that's trying to tell you what to do. So we're actually building on top of Giraffe. We're building our own algorithms. We're building our own interfaces."
Greystar has developed custom apps on the platform — tools built around their specific product types, their cost data, their design standards — that wouldn't exist if they were working with a closed, prescriptive tool. The flexibility to extend the platform is what makes it genuinely useful at Greystar's scale and complexity.
There's also a longer-term dimension to this. Greystar isn't a for-sale developer. They hold their buildings and manage them — the building that opens today has to perform equally well in 20 years. That changes what a digital model is for. McLuskey sees Giraffe as a collaboration platform that provides users with data, geometry and intelligence layers built in, along with algorithm tools to develop bespoke applications.
That's a different relationship with technology than most developers have. And it's one that's only possible if the platform is genuinely integrated across the business rather than being a design tool that gets archived when the building completes.
The takeaway for developers and investors
The Greystar approach suggests a few things worth considering for anyone deploying capital into residential development.
Speed and cost certainty are achievable earlier. The combination of advanced design (product resolved before site acquisition) and real-time cost feedback through the design process means that the gap between site control and investment decision can compress significantly. A week to construction drawings and a costed design is a meaningful advantage in competitive land markets.
Architect involvement correlates with better outcomes — when structured correctly. Greystar's model gives architects cost and value visibility from day one. That's not a concession to professional practice; it's a deliberate decision to put the people with the best design judgment in a position where they can also make commercially sound decisions. The buildings that come out of that process perform better.
Resident data should drive design, not policy compliance. The buildings Greystar is most proud of — and the ones that perform best operationally — break planning guidelines regularly. That's not an argument for ignoring regulation, but it is an argument for understanding the gap between what policy requires and what residents actually want, and designing intelligently within that space.
Mid-rise density is undervalued. The institutional preference for towers — driven partly by land economics, partly by planning frameworks — isn't aligned with where residents want to live or how communities thrive. Developers who are willing to explore denser mid-rise typologies may find better long-term operational performance than their tower-focused counterparts.
The AI story in property development is still being written. But Greystar's experience suggests the most valuable applications aren't the ones that make architects redundant. They're the ones that give architects — and by extension, the developers and investors working with them — better information, faster, and at the right point in the process.
Everything else, as McLuskey might say, is pointing in the wrong direction.