Goodbye, Phase 3! - hn Reader

High fidelity AI design deliverables kill 1/4 of the Double Diamond and create Design’s Jevon’s paradox

I spend a lot of my time thinking about how AI changes design practice, since I’m building an AI design tool. Here’s the most distilled version of my current perspective, which I put together for the talk I gave at Toyota Research back in June.

Start with the classic British Design Council double diamond, the most popular distillation of a complex iterative process that combines multi-variate constraint discovery (“The Problem”, left diamond) and a method for satisfying those constraints (“The Solution”, right diamond). This model is, at best, a so-so distillation, but it’s compact and popular so let’s go with it.

The classic double diamond

Scale the components so they more accurately represent the resources given to the tasks and you see that “Develop”, the part doing the detailed work of creating polished assets, uses by far most of the resources. It’s where most of the time is spent and it’s the thing that most non-designers and non-engineers consider to be design. The other work–the needs finding, the iterative evaluation–isn’t as visible and it’s culturally seen as ancillary to the design process.

IMO most effort in design goes to the Develop quadrant. There are obviously loops (it’s a lossy model, after all), but this quadrant is where the deliverables we associate with the core of design and engineering get made. To many non-designers (and designers) this quadrant *is* design.

AI makes that entire thing vanish. Poof. Deliverables arrive instantly “complete” and polished. Non-designer/non-engineer stakeholders think it replaces design because, from their perception of it, it does.

AI’s ability to generate high fidelity deliverables makes the Develop quadrant vanish, at least from the perspective of non-design stakeholders. Also, pragmatically, it does in fact largely vanish for many routine tasks.

Designers and engineers have two conniptions when this happens: one, because they know developing deliverables is only a fraction of the story; the second because many of them have, legitimately, defined their value and their job by their ability to execute in that quadrant.
Where that leaves Design is that those other quadrants still have to be done. They don’t disappear when the pixel pushing/code refactoring/CAD futzing is gone. However, now it’s much harder to explain what design is, to defend its value, and Design needs to find a new set of practices, and a new set of tools, to define it.

Go head, vacuum that thing, and make sure all the pile faces the same direction.

This is where Design’s Jevon paradox comes in. When a resource becomes much cheaper it doesn’t make existing processes become much cheaper, it changes the expectations from the system. Ruth Schwartz Cowan documented, in More Work for Mother, how home automation does not in fact reduce the amount of domestic labor, it changed the expectations for the results of that labor. Vacuum cleaners gave cover to architects to design high-maintenance houses with wall-to-wall white shag rugs and “conversation pits” that demanded, you guessed it, more work for mother as fitted carpet manufacturers made bank.

As I make my AI-first product design tool, I assume the third quadrant is gone. I don’t try to automate existing tools from that quadrant with AI automation gloss. It’s gone, 🫗, replaced by instant high-fidelity deliverables. In its place I’m building a workflow based on the other quadrants, so designers can use those instant high fidelity deliverables to think through problems better, to explore possibilities that would have been too expensive to explore before, to collaborate closer with stakeholders and customers, and to evaluate the results more effectively.