/9 min read
Working With AI on a Landed Project
In short
The specific, checkable version of a claim most companies will soon make in general terms. AI sits in five places on a KAPA project: testing massing against a plot's envelope in days rather than weeks, turning concepts into images early enough to choose between them, flagging budget drift against completed project data at drawing stage rather than at tender, reading monthly site reporting for early signs of delay, and making the project record searchable. Every decision is still a person's.
In this article
Topics
- Working with AI
- Process
- Planning
The claim worth making is the small one
By next year every design and build company in Singapore will say it uses AI. Almost none will say what for, because the specific version of the claim is harder to make and easier to check.
The honest position is stronger than the grand one. Here is where it actually sits in a landed project, stage by stage.
Before you buy: testing a plot
Landed housing is envelope-controlled, which means the shape of what can be built is fixed by the site before any design happens. Testing massing and envelope options against those constraints used to take a month of drawings.
It now takes days. That matters because it happens before the purchase decision, when walking away is still free.
D2 to D3: seeing more options
Turning a concept sketch into something photoreal quickly means you choose from what you can picture rather than from what you can imagine. More options are seen earlier, and the weak ones are discarded before anyone has invested a fortnight in them.
The drawings that go to authority are still drawn properly, by people, from the option you chose.
Budgeting and construction
Pattern-matching a live scheme against completed project data flags budget drift at drawing stage rather than at tender. Reading monthly site reporting for the patterns that precede a delay raises programme risk while there is still time to act.
Both are pattern recognition over a company's own history. Neither is a judgment.
The project record
Every drawing, decision, specification and approval, searchable. The question this answers is the one every client asks in month fourteen: what did we decide about the stair, and when?
That used to take a week. It now takes a minute. Every decision on your house is made by a person. What has changed is how quickly a person can get to the point of deciding well.
Key takeaways
- AI is used for five narrow, checkable tasks, not for design.
- Massing tested against plot constraints in days shifts the work before purchase.
- Budget drift is flagged at drawing stage rather than at tender.
- A searchable project record answers month-fourteen questions in a minute.
First published in the KAPA Journal, written with KAPA Design Collective, the practice that designs and builds our residences.
Published by TMB Development at tmbdevelopment.com/journal/working-with-ai-on-a-landed-project. Quoting a passage with a link back is welcome; republishing it is not — see the terms.
