Why capital allocators need a platform, not a tool

Software is a tiny, inexpensive machine compared to the deals it shapes. The developer with the better system wins better deals.

Say you need to underwrite a site. You can do it many ways. A rectangle and some napkin maths might be enough. Or you may need to show key features — sketch a core position, a road entrance — to prove the deal works, to yourself and to others. You may even need unit layouts so an agent can sense the product.

Giraffe is designed to accommodate all of these. You can see it in the video.

Sometimes this is frustrating because Giraffe kind of matches existing BIM or CAD tools, but then frustratingly changes things. There are almost always reasons for the changes — it is a hybrid tool for DMs, designers, and policy makers, so there are workflows that are completely novel, and need novel inputs.

But we have worked very hard to keep it flexible. More than that, Giraffe is built so you can shape your own process. So if you're a student housing developer working to a strict module and you want the tool rigid, unbending, efficient — Giraffe can do that. Or if you build bespoke luxury houses, where curves and custom concrete are expected, Giraffe can do that too. And not just in terms of modelling, but in building automated workflows.

This, we find, is the hard part: looking at a workflow. Identifying the jobs to do, how long they take, and which of them could in principle be automated. Once that is done we just build the automation.

The relative cost of software vs. deals

The urgency is in the relative difference in cost between the software and the deals. The algorithm at the end of the video took about three hours to build. It will now be deployed widely to do deals much larger than $20m, some close to $200m or even more.

Software is a tiny, inexpensive machine compared to that, but it is a force multiplier. The developer with the better system wins better deals.

And so you should think of the software as plastic and changeable. There are always objections: the tool doesn't do X, or Y. Often it doesn't, but the better question is what can we make it do, and in what time. Imagine 40 hours spent on the algorithm in the video — even at $500 an hour, that's $20,000 of work shaping a system that will keep paying back across every deal it touches.

Visualising the cumulative value of automation

To visualise the cumulative value of systematically automating work I built the animation below. Have a play with the parameters to get order-of-magnitude savings — bespoke design (top) versus a systemised, kit-of-parts approach (bottom), running through the same messy sites.